1-library-matching-improvement.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Spec2Vec - Find similar compounds for unknown substances\n",
    "One very important use case for a spectral similariy measure is libarary matching.  \n",
    "This means matching previously unseen/unknown spectra to known, annotated sepctra in a larger library."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import sys\n",
    "import gensim\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "ROOT = os.path.dirname(os.getcwd())\n",
    "#path_data = os.path.join(ROOT, 'data')\n",
    "path_data = 'C:\\\\Users\\\\joris\\\\Documents\\\\eScience_data\\\\data'\n",
    "sys.path.insert(0, ROOT)\n",
    "sys.path.insert(0, \"C:\\\\Users\\\\joris\\\\Documents\\\\eScience_data\\\\spec2vec_gnps_data_analysis\\\\custom_functions\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Import pre-processed dataset \"AllPositive\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C:\\Users\\joris\\Documents\\eScience_data\\data\\gnps_positive_ionmode_cleaned_by_matchms_and_lookups.pickle\n",
      "number of spectra: 112956\n"
     ]
    }
   ],
   "source": [
    "import pickle\n",
    "from matchms.importing import load_from_json\n",
    "outfile = os.path.join(path_data, 'gnps_positive_ionmode_cleaned_by_matchms_and_lookups.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        spectrums = pickle.load(inf)\n",
    "else:\n",
    "    filename = os.path.join(path_data,'gnps_positive_ionmode_cleaned_by_matchms_and_lookups.json')\n",
    "    spectrums = load_from_json(filename)\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(spectrums, outf)\n",
    "\n",
    "print(\"number of spectra:\", len(spectrums))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C:\\Users\\joris\\Documents\\eScience_data\\data\\gnps_positive_ionmode_cleaned_by_matchms_and_lookups.pickle\n",
      "loading took 29.74 s\n"
     ]
    }
   ],
   "source": [
    "#loading data as pickled object goes a lot quicker\n",
    "import time\n",
    "print(outfile)\n",
    "start = time.time()\n",
    "with open(outfile, 'rb') as inf:\n",
    "        spectrums = pickle.load(inf)\n",
    "end = time.time()\n",
    "print('loading took {:.2f} s'.format(end-start))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'number of spectra in respective bin')"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "number_of_peaks = [len(spec.peaks) for spec in spectrums]\n",
    "\n",
    "plt.figure(figsize=(12,7))\n",
    "hist = plt.hist(number_of_peaks, np.arange(0,2000,20))\n",
    "plt.xlabel(\"number of peaks in spectrum\")\n",
    "plt.ylabel(\"number of spectra in respective bin\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Post-process data\n",
    "We here do different post-processing for either the \"classical\" scores (Cosine, modified Cosine) or Spe2Vec. The reason for this is mostly to make sure that we do full justice to the classical similarity measures.\n",
    "\n",
    "#### Post-processing for classical scores\n",
    "Cosine and modified cosine scores are build on multiplying intensities of matching peaks. Small peaks (<0.01) hence contribute very little to the final scores and are left out to avoid excessive computation times. \n",
    "+ normalize peaks (maximum intensity to 1)\n",
    "+ remove peaks outside [0, 1000] m/z window\n",
    "+ remove spectra with < 10 peaks\n",
    "+ remove peaks with intensities < 0.01 of maximum intensity.\n",
    "\n",
    "#### Post-processing for Spec2Vec\n",
    "Spec2Vec similarity scores rely on creating a ``document vector`` for each spectrum. For the underlying word2vec model we want the documents (=spectra) to be more homogeneous in their number of unique words. Assuming that larger compounds will on average break down into a higher number of meaningful fragment peaks we reduce the document size of each spectrum according to its parent mass.\n",
    "+ normalize peaks (maximum intensity to 1)\n",
    "+ remove peaks outside [0, 1000] m/z window\n",
    "+ remove spectra with < 10 peaks\n",
    "+ reduce number of peaks to maximum of 0.5 * parent mass\n",
    "+ remove peaks with intensities < 0.001 of maximum intensity (unless this brings number of peaks to less than 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "95320 remaining spectra.\n"
     ]
    }
   ],
   "source": [
    "from matchms.filtering import normalize_intensities\n",
    "from matchms.filtering import require_minimum_number_of_peaks\n",
    "from matchms.filtering import select_by_mz\n",
    "from matchms.filtering import select_by_relative_intensity\n",
    "from matchms.filtering import reduce_to_number_of_peaks\n",
    "from matchms.filtering import add_losses\n",
    "\n",
    "def post_process_s2v(s):\n",
    "    s = normalize_intensities(s)\n",
    "    s = select_by_mz(s, mz_from=0, mz_to=1000)\n",
    "    s = require_minimum_number_of_peaks(s, n_required=10)\n",
    "    s = reduce_to_number_of_peaks(s, n_required=10, ratio_desired=0.5)\n",
    "    if s is None:\n",
    "        return None\n",
    "    s_remove_low_peaks = select_by_relative_intensity(s, intensity_from=0.001)\n",
    "    if len(s_remove_low_peaks.peaks) >= 10:\n",
    "        s = s_remove_low_peaks\n",
    "        \n",
    "    s = add_losses(s, loss_mz_from=5.0, loss_mz_to=200.0)\n",
    "    return s\n",
    "\n",
    "# apply post processing steps to the data\n",
    "spectrums_s2v = [post_process_s2v(s) for s in spectrums]\n",
    "\n",
    "# omit spectrums that didn't qualify for analysis\n",
    "spectrums_s2v = [s for s in spectrums_s2v if s is not None]\n",
    "\n",
    "print(\"{} remaining spectra.\".format(len(spectrums_s2v)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'number of spectra in respective bin')"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "number_of_peaks = [len(spec.peaks) for spec in spectrums_s2v]\n",
    "\n",
    "plt.figure(figsize=(12,7))\n",
    "hist = plt.hist(number_of_peaks, np.arange(0,2000,20))\n",
    "plt.xlabel(\"number of peaks in spectrum\")\n",
    "plt.ylabel(\"number of spectra in respective bin\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1081"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.max(number_of_peaks)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Inspect annotations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "Inchikeys = []\n",
    "for spec in spectrums_s2v:\n",
    "    Inchikeys.append(spec.get(\"inchikey\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['IYDKWWDUBYWQGF-NNAZGLEUSA-N',\n",
       " 'KNGPFNUOXXLKCN-ZNCJFREWSA-N',\n",
       " 'WXDBUBIFYCCNLE-NSCMQRKRSA-N',\n",
       " 'CGZKSPLDUIRCIO-RPCRKUJJSA-N',\n",
       " '',\n",
       " 'USXIYWCPCGVOKF-LERJCCFDSA-N',\n",
       " '',\n",
       " 'WCNJVJCYRBJSLC-BCJYPDSRSA-N',\n",
       " 'USXIYWCPCGVOKF-NOENWEJRSA-N',\n",
       " 'WCNJVJCYRBJSLC-BCJYPDSRSA-N']"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Inchikeys[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "NEGQHKSYEYVFTD    432\n",
       "SULIDBRAXVDKBU    426\n",
       "IQGPMZRCLCCXAG    308\n",
       "WTJKGGKOPKCXLL    295\n",
       "IIZPXYDJLKNOIY    235\n",
       "RWKUXQNLWDTSLO    234\n",
       "QIBZFHLFHCIUOT    225\n",
       "ZAYXPDDGEIJGGW    212\n",
       "QJWDAOSZZYVBJZ    210\n",
       "KILNVBDSWZSGLL    205\n",
       "RFVFQQWKPSOBED    202\n",
       "IESVDEZGAHUQJU    197\n",
       "LFUDDCMNKWEORN    195\n",
       "JLPULHDHAOZNQI    182\n",
       "LSOWKZULVQWMLY    180\n",
       "XGGMHQYOVYWRLV    178\n",
       "LLHISNQVRRYJGL    175\n",
       "JFISYPWOVQNHLS    174\n",
       "SRIGHEHXEGELQJ    166\n",
       "ACTIUHUUMQJHFO    163\n",
       "PZNPLUBHRSSFHT    162\n",
       "YLWSJLLZUHSIEA    158\n",
       "GPWHCUUIQMGELX    150\n",
       "BLZVZPYMHLXLHG    148\n",
       "QEDPUVGSSDPBMD    146\n",
       "IGZPHNNYPPAPLA    142\n",
       "CITHEXJVPOWHKC    141\n",
       "SXNXGNVZTLZDHE    131\n",
       "YEJYLHKQOBOSCP    131\n",
       "AXZGUCXCTZMPTR    128\n",
       "dtype: int64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "inchikeys_pd = pd.Series([x for x in Inchikeys if x])\n",
    "inchikeys_pd.str[:14].value_counts()[:30]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'Number of inchikeys with respective number of occurences')"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(12,6))\n",
    "plt.hist(inchikeys_pd.str[:14].value_counts(), np.linspace(0, 60, 60))\n",
    "plt.xlabel('Number of occurences of inchikey')\n",
    "plt.ylabel('Number of inchikeys with respective number of occurences')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>inchikey14</th>\n",
       "      <th>occurences</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>NEGQHKSYEYVFTD</td>\n",
       "      <td>432</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>SULIDBRAXVDKBU</td>\n",
       "      <td>426</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>IQGPMZRCLCCXAG</td>\n",
       "      <td>308</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>WTJKGGKOPKCXLL</td>\n",
       "      <td>295</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>IIZPXYDJLKNOIY</td>\n",
       "      <td>235</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>RWKUXQNLWDTSLO</td>\n",
       "      <td>234</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>QIBZFHLFHCIUOT</td>\n",
       "      <td>225</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>ZAYXPDDGEIJGGW</td>\n",
       "      <td>212</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>QJWDAOSZZYVBJZ</td>\n",
       "      <td>210</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>KILNVBDSWZSGLL</td>\n",
       "      <td>205</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>RFVFQQWKPSOBED</td>\n",
       "      <td>202</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>IESVDEZGAHUQJU</td>\n",
       "      <td>197</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>LFUDDCMNKWEORN</td>\n",
       "      <td>195</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>JLPULHDHAOZNQI</td>\n",
       "      <td>182</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>LSOWKZULVQWMLY</td>\n",
       "      <td>180</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        inchikey14  occurences\n",
       "0   NEGQHKSYEYVFTD         432\n",
       "1   SULIDBRAXVDKBU         426\n",
       "2   IQGPMZRCLCCXAG         308\n",
       "3   WTJKGGKOPKCXLL         295\n",
       "4   IIZPXYDJLKNOIY         235\n",
       "5   RWKUXQNLWDTSLO         234\n",
       "6   QIBZFHLFHCIUOT         225\n",
       "7   ZAYXPDDGEIJGGW         212\n",
       "8   QJWDAOSZZYVBJZ         210\n",
       "9   KILNVBDSWZSGLL         205\n",
       "10  RFVFQQWKPSOBED         202\n",
       "11  IESVDEZGAHUQJU         197\n",
       "12  LFUDDCMNKWEORN         195\n",
       "13  JLPULHDHAOZNQI         182\n",
       "14  LSOWKZULVQWMLY         180"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "suitable_inchikeys = pd.DataFrame(inchikeys_pd.str[:14].value_counts()[inchikeys_pd.str[:14].value_counts().values >= 5])\n",
    "suitable_inchikeys.reset_index(level=suitable_inchikeys.index.names, inplace=True)\n",
    "suitable_inchikeys.columns = (['inchikey14', 'occurences'])\n",
    "\n",
    "# Important: sort values to make it reproducible (same occurences have random order otherwise!)\n",
    "suitable_inchikeys = suitable_inchikeys.sort_values(['occurences', 'inchikey14'], ascending=False)\n",
    "suitable_inchikeys.head(15)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of spectra with inchikeys that exist >= 5 times: 4079\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of spectra with inchikeys that exist >= 5 times:\", suitable_inchikeys.shape[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Randomly select 1000 inchikeys that exist >=5  times in the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[  33 3233 1556 2547  457 3598  857 1878  705  803 1642  495 2866 1034\n",
      " 1517  564 2826  568 2575 3956  670 3317 1498 2971   96]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array(['SRRQPVVYXBTRQK', 'JUZYLCPPVHEVSV', 'UZXMLGUMBQQVME',\n",
       "       'PDAKXMIQFUHWQC', 'JFVKWCYZKMUTLH', 'HQPCSDADVLFHHO',\n",
       "       'HUPGTAGQEXENPN', 'YPKUMLKVFXFYOT', 'KWIUHFFTVRNATP',\n",
       "       'QJVXKWHHAMZTBY', 'PGERTGWKXFAEFR', 'MIHLTJZHRJACQQ',\n",
       "       'YFAGHNZHGGCZAX', 'LKWWJGGLULNRBP', 'XWTYSIMOBUGWOL',\n",
       "       'BPEXJHGGARTCIR', 'FHHVIBPVBBRLOR', 'YFPYXTNSQOUHPS',\n",
       "       'LTLYEAJONXGNFG', 'INOGLHRUEYDAHX', 'VGOJYSUPEJWUNN',\n",
       "       'BPICBUSOMSTKRF', 'ZQHJXKYYELWEOK', 'MXHRCPNRJAMMIM',\n",
       "       'QXMHHXQBBKDSSL'], dtype=object)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "num_spectra = 1000\n",
    "\n",
    "np.random.seed(42) # to make it reproducible\n",
    "selection = np.random.choice(suitable_inchikeys.shape[0], num_spectra, replace=False)\n",
    "print(selection[:25])\n",
    "selected_inchikeys = suitable_inchikeys['inchikey14'].values[selection]\n",
    "selected_inchikeys[:25]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3079\n",
      "Number of old selected inchikeys in new list: 0\n",
      "[  63 2775 2016  331 2739  594 2244  346 2312 1472 2785  194 2370 2032\n",
      " 2538  246  256  486  978 1961  144  801  744  457  416] 1000\n",
      "Overlap between new and old list: 0\n"
     ]
    }
   ],
   "source": [
    "#Randomly select another 1000 inchikeys that exist >=5 times in the dataset and don't occur in the first list selected_inchikeys\n",
    "new_suitable_inchikeys = suitable_inchikeys.copy()\n",
    "drop_selection = suitable_inchikeys.index[selection] #get index names of first selection\n",
    "new_suitable_inchikeys.drop(drop_selection, inplace = True) #remove first selection\n",
    "print(new_suitable_inchikeys.shape[0])\n",
    "print('Number of old selected inchikeys in new list:',len(set(new_suitable_inchikeys['inchikey14']) & set(selected_inchikeys)))\n",
    "\n",
    "np.random.seed(42) # to make it reproducible\n",
    "new_selection = np.random.choice(new_suitable_inchikeys.shape[0], num_spectra, replace=False)\n",
    "print(new_selection[:25], len(new_selection))\n",
    "new_selected_inchikeys = new_suitable_inchikeys['inchikey14'].values[new_selection]\n",
    "print('Overlap between new and old list:', len(set(new_selected_inchikeys) & set(selected_inchikeys)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2079\n",
      "Number of old selected inchikeys in new list: 0\n",
      "[1362 2004  607   29  429  861  109  693  432   51 1969 1389 1532  354\n",
      "  188 1972  251 1229  571 1916  561  342 1548 2031 1292] 1000\n",
      "Overlap between new and old list: 0\n"
     ]
    }
   ],
   "source": [
    "#Randomly select another 1000 inchikeys that exist >=5 times in the dataset and don't occur in the first two lists\n",
    "new2_suitable_inchikeys = new_suitable_inchikeys.copy()\n",
    "new2_drop_selection = new_suitable_inchikeys.index[new_selection] #get index names of first selection\n",
    "new2_suitable_inchikeys.drop(new2_drop_selection, inplace = True) #remove first selection\n",
    "print(new2_suitable_inchikeys.shape[0])\n",
    "print('Number of old selected inchikeys in new list:',len(set(new2_suitable_inchikeys['inchikey14']) & set(new_selected_inchikeys)))\n",
    "\n",
    "np.random.seed(42) # to make it reproducible\n",
    "new2_selection = np.random.choice(new2_suitable_inchikeys.shape[0], num_spectra, replace=False)\n",
    "print(new2_selection[:25], len(new2_selection))\n",
    "new2_selected_inchikeys = new2_suitable_inchikeys['inchikey14'].values[new2_selection]\n",
    "print('Overlap between new and old list:', len(set(new2_selected_inchikeys) & set(new_selected_inchikeys)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1079\n",
      "Number of old selected inchikeys in new list: 0\n",
      "[ 423  548  730  588  830  616  989  991  950 1004  728  581  634  158\n",
      "  365  448  333  495  910  937   56   70  485  388  528] 1000\n",
      "Overlap between new and old list: 0\n"
     ]
    }
   ],
   "source": [
    "#Randomly select another 1000 inchikeys that exist >=5 times in the dataset and don't occur in the first two lists\n",
    "new3_suitable_inchikeys = new2_suitable_inchikeys.copy()\n",
    "new3_drop_selection = new2_suitable_inchikeys.index[new2_selection] #get index names of first selection\n",
    "new3_suitable_inchikeys.drop(new3_drop_selection, inplace = True) #remove first selection\n",
    "print(new3_suitable_inchikeys.shape[0])\n",
    "print('Number of old selected inchikeys in new list:',len(set(new3_suitable_inchikeys['inchikey14']) & set(new2_selected_inchikeys)))\n",
    "\n",
    "np.random.seed(42) # to make it reproducible\n",
    "new3_selection = np.random.choice(new3_suitable_inchikeys.shape[0], num_spectra, replace=False)\n",
    "print(new3_selection[:25], len(new3_selection))\n",
    "new3_selected_inchikeys = new3_suitable_inchikeys['inchikey14'].values[new3_selection]\n",
    "print('Overlap between new and old list:', len(set(new3_selected_inchikeys) & set(new2_selected_inchikeys)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Randomly pick 1000 inchikeys (spectra) that are unqiue in the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>inchikey14</th>\n",
       "      <th>occurences</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3285</th>\n",
       "      <td>ZZYHCCDMBJTROG</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>67</th>\n",
       "      <td>ZZTCNNZHOWDRPS</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4149</th>\n",
       "      <td>ZZQHNBGRWRQWFI</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3168</th>\n",
       "      <td>ZZPAWQYZQVUVHX</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3736</th>\n",
       "      <td>ZZNVCZGRNCQHCQ</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1629</th>\n",
       "      <td>ZZHLYYDVIOPZBE</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4919</th>\n",
       "      <td>ZZAJQOPSWWVMBI</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2303</th>\n",
       "      <td>ZYYNEJWFGGVJQZ</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1415</th>\n",
       "      <td>ZYVYPNZFOCZLEM</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5309</th>\n",
       "      <td>ZYUDDSCUGFXAKK</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>822</th>\n",
       "      <td>ZYTHYMZQQBWDDS</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1031</th>\n",
       "      <td>ZYGYNVJVDLQWDG</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1222</th>\n",
       "      <td>ZYGSNVAGWKWKKI</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1789</th>\n",
       "      <td>ZYEMGPIYFIJGTP</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1905</th>\n",
       "      <td>ZXZPFGGUUZDXND</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          inchikey14  occurences\n",
       "3285  ZZYHCCDMBJTROG           1\n",
       "67    ZZTCNNZHOWDRPS           1\n",
       "4149  ZZQHNBGRWRQWFI           1\n",
       "3168  ZZPAWQYZQVUVHX           1\n",
       "3736  ZZNVCZGRNCQHCQ           1\n",
       "1629  ZZHLYYDVIOPZBE           1\n",
       "4919  ZZAJQOPSWWVMBI           1\n",
       "2303  ZYYNEJWFGGVJQZ           1\n",
       "1415  ZYVYPNZFOCZLEM           1\n",
       "5309  ZYUDDSCUGFXAKK           1\n",
       "822   ZYTHYMZQQBWDDS           1\n",
       "1031  ZYGYNVJVDLQWDG           1\n",
       "1222  ZYGSNVAGWKWKKI           1\n",
       "1789  ZYEMGPIYFIJGTP           1\n",
       "1905  ZXZPFGGUUZDXND           1"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "suitable_inchikeys = pd.DataFrame(inchikeys_pd.str[:14].value_counts()[inchikeys_pd.str[:14].value_counts().values >= 5])\n",
    "unique_inchikeys = pd.DataFrame(inchikeys_pd.str[:14].value_counts()[inchikeys_pd.str[:14].value_counts().values == 1])\n",
    "\n",
    "unique_inchikeys.reset_index(level=unique_inchikeys.index.names, inplace=True)\n",
    "unique_inchikeys.columns = (['inchikey14', 'occurences'])\n",
    "\n",
    "# Important: sort values to make it reproducible (same occurences have random order otherwise!)\n",
    "unique_inchikeys = unique_inchikeys.sort_values(['occurences', 'inchikey14'], ascending=False)\n",
    "unique_inchikeys.head(15)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5352\n"
     ]
    }
   ],
   "source": [
    "print(len(unique_inchikeys))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[4114 1650  401 4221 1684   33  290 3265 5324 3235 4213 3403 5004 3699\n",
      " 1075  925 4074 4446 2642 2972  810 3204 3872 2707 2651]\n",
      "['GQIJYUMTOUBHSH' 'RVWQXFWAUMMLKE' 'YGULWPYYGQCFMP' 'GDPHPXYFLPDZGH'\n",
      " 'RRNSPXUFTKJIEZ' 'ZWNKGUOHTAKRBN' 'YTNXARXLPJGCHV' 'KMRZMOYMAAVHNA'\n",
      " 'ADDCNOCQPWDJSR' 'KQFUXLQBMQGNRT' 'GEJNVNUFFYATKR' 'JVRVKOOOXHGJKI'\n",
      " 'BOVRCQYBOHNUIF' 'INOKSQLHQGQUNF' 'VGGCLGHLLZKCSW' 'VYGQXRZAHIZHQV'\n",
      " 'GUOQUXNJZHGPQF' 'FAZKBTFQSATIIN' 'NISZLIJBKRGIAU' 'LVUQCTGSDJLWCE'\n",
      " 'WMAITHDYVBQITD' 'KSZJPQHIPKEEMF' 'HSLDNQNDHKZKNZ' 'NBLBCGUCPBXKOV'\n",
      " 'NHNODHRSCRALBF']\n",
      "1000\n"
     ]
    }
   ],
   "source": [
    "num_spectra = 1000\n",
    "\n",
    "np.random.seed(42) # to make it reproducible\n",
    "unique_selection = np.random.choice(unique_inchikeys.shape[0], num_spectra, replace=False)\n",
    "print(unique_selection[:25])\n",
    "selected_unique_inchikeys = unique_inchikeys['inchikey14'].values[unique_selection]\n",
    "print(selected_unique_inchikeys[:25])\n",
    "\n",
    "#get the spectrum that belongs with the unique chosen inchikeys\n",
    "selected_unique_spectra = []\n",
    "inchikeys_pd = pd.Series([x for x in Inchikeys]) #include all even empty ones to get the IDs right!\n",
    "\n",
    "np.random.seed(42) # to make it reproducible\n",
    "for inchikey in selected_unique_inchikeys:\n",
    "    matches = inchikeys_pd[inchikeys_pd.str[:14] == inchikey].index.values\n",
    "    selected_unique_spectra.append(int(matches[0]))\n",
    "print(len(selected_unique_spectra))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4352\n",
      "Number of old selected inchikeys in new list: 0\n",
      "[ 505 1611 3107  511 1467  410 3678 2677 3370 4235 3853 3217  149 4332\n",
      " 1295 2457 2589 1261  287 3038   17  755  305 3276 4084] 1000\n",
      "Overlap between new and old list: 0\n",
      "1000\n"
     ]
    }
   ],
   "source": [
    "#Randomly select another 1000 inchikeys that are unique in the dataset and don't occur in the first list selected_unique_inchikeys\n",
    "new_unique_inchikeys = unique_inchikeys.copy()\n",
    "drop_selection = unique_inchikeys.index[unique_selection] #get index names of first selection\n",
    "new_unique_inchikeys.drop(drop_selection, inplace = True) #remove first selection\n",
    "print(new_unique_inchikeys.shape[0])\n",
    "print('Number of old selected inchikeys in new list:',len(set(new_unique_inchikeys['inchikey14']) & set(selected_unique_inchikeys)))\n",
    "\n",
    "np.random.seed(42) # to make it reproducible\n",
    "new_unique_selection = np.random.choice(new_unique_inchikeys.shape[0], num_spectra, replace=False)\n",
    "print(new_unique_selection[:25], len(new_unique_selection))\n",
    "new_selected_unique_inchikeys = new_unique_inchikeys['inchikey14'].values[new_unique_selection]\n",
    "print('Overlap between new and old list:', len(set(new_selected_unique_inchikeys) & set(selected_unique_inchikeys)))\n",
    "\n",
    "#get the spectrum that belongs with the unique chosen inchikeys\n",
    "new_selected_unique_spectra = []\n",
    "inchikeys_pd = pd.Series([x for x in Inchikeys]) #include all even empty ones to get the IDs right!\n",
    "\n",
    "np.random.seed(42) # to make it reproducible\n",
    "for inchikey in new_selected_unique_inchikeys:\n",
    "    matches = inchikeys_pd[inchikeys_pd.str[:14] == inchikey].index.values\n",
    "    new_selected_unique_spectra.append(int(matches[0]))\n",
    "print(len(new_selected_unique_spectra))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Randomly pick one spectra for each of the chosen inchikeys"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "selected_spectra = []\n",
    "inchikeys_pd = pd.Series([x for x in Inchikeys]) #include all even empty ones to get the IDs right!\n",
    "\n",
    "np.random.seed(42) # to make it reproducible\n",
    "for inchikey in selected_inchikeys:\n",
    "    matches = inchikeys_pd[inchikeys_pd.str[:14] == inchikey].index.values\n",
    "    selected_spectra.append(int(np.random.choice(matches,1)[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2000"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#add unique dataset together with old test set of 1000 spectra >5 occurrence\n",
    "old_and_unique_test_set = selected_spectra + selected_unique_spectra\n",
    "len(old_and_unique_test_set)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[52728, 6129, 78135, 62556, 73314, 21762, 12492, 88980, 63188, 74380]"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "selected_spectra[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CCMSLIB00003128874 SRRQPVVYXBTRQK\n",
      "CCMSLIB00000005841 JUZYLCPPVHEVSV\n",
      "CCMSLIB00004701389 UZXMLGUMBQQVME\n",
      "CCMSLIB00000215466 PDAKXMIQFUHWQC\n",
      "CCMSLIB00004695353 JFVKWCYZKMUTLH\n",
      "CCMSLIB00003136589 HQPCSDADVLFHHO\n",
      "CCMSLIB00000850213 HUPGTAGQEXENPN\n",
      "CCMSLIB00004715739 YPKUMLKVFXFYOT\n",
      "CCMSLIB00000217500 KWIUHFFTVRNATP\n",
      "CCMSLIB00004696689 QJVXKWHHAMZTBY\n"
     ]
    }
   ],
   "source": [
    "# Show gnps IDs for selected spectra\n",
    "for ID in selected_spectra[:10]:\n",
    "    print(spectrums_s2v[ID].get(\"spectrumid\"),\n",
    "         spectrums_s2v[ID].get(\"inchikey\")[:14])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Pick one spectrum for each of the new chosen inchikeys\n",
    "new_selected_spectra = []\n",
    "\n",
    "np.random.seed(42) # to make it reproducible\n",
    "for inchikey in new_selected_inchikeys:\n",
    "    matches = inchikeys_pd[inchikeys_pd.str[:14] == inchikey].index.values\n",
    "    new_selected_spectra.append(int(np.random.choice(matches,1)[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "#add second unique test set together with new test set of 1000 spectra >5 occurrence\n",
    "new_and_unique2_test_set = new_selected_spectra + new_selected_unique_spectra"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Pick one spectrum for each of the new chosen inchikeys\n",
    "new2_selected_spectra = []\n",
    "new3_selected_spectra = []\n",
    "\n",
    "np.random.seed(42) # to make it reproducible\n",
    "for inchikey in new2_selected_inchikeys:\n",
    "    matches = inchikeys_pd[inchikeys_pd.str[:14] == inchikey].index.values\n",
    "    new2_selected_spectra.append(int(np.random.choice(matches,1)[0]))\n",
    "\n",
    "np.random.seed(42) # to make it reproducible\n",
    "for inchikey in new3_selected_inchikeys:\n",
    "    matches = inchikeys_pd[inchikeys_pd.str[:14] == inchikey].index.values\n",
    "    new3_selected_spectra.append(int(np.random.choice(matches,1)[0]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train new model on AllPositive dataset WITHOUT selected spectra"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "from spec2vec import Spec2VecParallel\n",
    "from spec2vec import SpectrumDocument\n",
    "from spec2vec.model_building import train_new_word2vec_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 242,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C:\\Users\\joris\\Documents\\eScience_data\\data\\documents_library_s2v.pickle\n"
     ]
    }
   ],
   "source": [
    "outfile = os.path.join(path_data, 'documents_library_s2v.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        documents_library_s2v = pickle.load(inf)\n",
    "else:\n",
    "    documents_library_s2v = [SpectrumDocument(s, n_decimals=2) for ind, s in enumerate(spectrums_s2v) if ind not in selected_spectra]\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(documents_library_s2v, outf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 241,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C:\\Users\\joris\\Documents\\eScience_data\\data\\new_documents_library_s2v.pickle\n"
     ]
    }
   ],
   "source": [
    "outfile = os.path.join(path_data, 'new_documents_library_s2v.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        new_documents_library_s2v = pickle.load(inf)\n",
    "else:\n",
    "    new_documents_library_s2v = [SpectrumDocument(s, n_decimals=2) for ind, s in enumerate(spectrums_s2v) if ind not in new_selected_spectra]\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(new_documents_library_s2v, outf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C:\\Users\\joris\\Documents\\eScience_data\\data\\old_and_unique_documents_library_s2v.pickle\n"
     ]
    }
   ],
   "source": [
    "outfile = os.path.join(path_data, 'old_and_unique_documents_library_s2v.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        old_and_unique_documents_library_s2v = pickle.load(inf)\n",
    "else:\n",
    "    old_and_unique_documents_library_s2v = [SpectrumDocument(s, n_decimals=2) for ind, s in enumerate(spectrums_s2v) if ind not in old_and_unique_test_set]\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(old_and_unique_documents_library_s2v, outf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C:\\Users\\joris\\Documents\\eScience_data\\data\\new_and_unique2_documents_library_s2v.pickle\n"
     ]
    }
   ],
   "source": [
    "outfile = os.path.join(path_data, 'new_and_unique2_documents_library_s2v.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        new_and_unique2_documents_library_s2v = pickle.load(inf)\n",
    "else:\n",
    "    new_and_unique2_documents_library_s2v = [SpectrumDocument(s, n_decimals=2) for ind, s in enumerate(spectrums_s2v) if ind not in new_and_unique2_test_set]\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(new_and_unique2_documents_library_s2v, outf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "new2_documents_library_s2v = [SpectrumDocument(s, n_decimals=2) for ind, s in enumerate(spectrums_s2v) if ind not in new2_selected_spectra]\n",
    "new3_documents_library_s2v = [SpectrumDocument(s, n_decimals=2) for ind, s in enumerate(spectrums_s2v) if ind not in new3_selected_spectra]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "94320 95320\n",
      "94320 95320\n"
     ]
    }
   ],
   "source": [
    "print(len(documents_library_s2v), len(spectrums_s2v))  # 1000 should be removed\n",
    "print(len(new_documents_library_s2v), len(spectrums_s2v))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "#from spec2vec.model_building import train_new_word2vec_model\n",
    "\n",
    "#path_models = os.path.join(path_data, \"trained_models\")\n",
    "#model_file = os.path.join(path_models, \"spec2vec_librarymatching.model\")\n",
    "\n",
    "#iterations = [1, 3, 5, 10, 15, 20]\n",
    "\n",
    "# Train model with default parameters\n",
    "#model = train_new_word2vec_model(documents_library_s2v, iterations, model_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "#train new model on allpositive without new selected spectra\n",
    "\n",
    "#from spec2vec.model_building import train_new_word2vec_model\n",
    "\n",
    "#path_models = os.path.join(path_data, \"trained_models\")\n",
    "#new_model_file = os.path.join(path_models, \"spec2vec_librarymatchingnew.model\")\n",
    "\n",
    "#iterations = [1, 3, 5, 10, 15, 20]\n",
    "\n",
    "# Train model with default parameters\n",
    "#model = train_new_word2vec_model(new_documents_library_s2v, iterations, new_model_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  Epoch 1 of 15.Change in loss after epoch 1: 5733627.0\n",
      "  Epoch 2 of 15.Change in loss after epoch 2: 4333161.0\n",
      "  Epoch 3 of 15.Change in loss after epoch 3: 3833238.0\n",
      "  Epoch 4 of 15.Change in loss after epoch 4: 3488840.0\n",
      "  Epoch 5 of 15.Change in loss after epoch 5: 2988288.0\n",
      "  Epoch 6 of 15.Change in loss after epoch 6: 2856058.0\n",
      "  Epoch 7 of 15.Change in loss after epoch 7: 2714448.0\n",
      "  Epoch 8 of 15.Change in loss after epoch 8: 2641868.0\n",
      "  Epoch 9 of 15.Change in loss after epoch 9: 2599070.0\n",
      "  Epoch 10 of 15.Change in loss after epoch 10: 2443018.0\n",
      "  Epoch 11 of 15.Change in loss after epoch 11: 1529916.0\n",
      "  Epoch 12 of 15.Change in loss after epoch 12: 1528688.0\n",
      "  Epoch 13 of 15.Change in loss after epoch 13: 1493040.0\n",
      "  Epoch 14 of 15.Change in loss after epoch 14: 1482232.0\n",
      "  Epoch 15 of 15.Change in loss after epoch 15: 1452728.0\n",
      "Saving model with name: C:\\Users\\joris\\Documents\\eScience_data\\data\\trained_models\\spec2vec_librarymatchingnew2.model\n"
     ]
    }
   ],
   "source": [
    "#train new model on allpositive without new selected spectra\n",
    "\n",
    "#path_models = os.path.join(path_data, \"trained_models\")\n",
    "#new2_model_file = os.path.join(path_models, \"spec2vec_librarymatchingnew2.model\")\n",
    "\n",
    "#iterations = [15]\n",
    "\n",
    "# Train model with default parameters\n",
    "#new2_model = train_new_word2vec_model(new2_documents_library_s2v, iterations, new2_model_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  Epoch 1 of 15.Change in loss after epoch 1: 5815822.5\n",
      "  Epoch 2 of 15.Change in loss after epoch 2: 4332776.5\n",
      "  Epoch 3 of 15.Change in loss after epoch 3: 3774687.0\n",
      "  Epoch 4 of 15.Change in loss after epoch 4: 3533044.0\n",
      "  Epoch 5 of 15.Change in loss after epoch 5: 2944196.0\n",
      "  Epoch 6 of 15.Change in loss after epoch 6: 2837714.0\n",
      "  Epoch 7 of 15.Change in loss after epoch 7: 2680200.0\n",
      "  Epoch 8 of 15.Change in loss after epoch 8: 2607796.0\n",
      "  Epoch 9 of 15.Change in loss after epoch 9: 2579052.0\n",
      "  Epoch 10 of 15.Change in loss after epoch 10: 2487492.0\n",
      "  Epoch 11 of 15.Change in loss after epoch 11: 1563744.0\n",
      "  Epoch 12 of 15.Change in loss after epoch 12: 1518200.0\n",
      "  Epoch 13 of 15.Change in loss after epoch 13: 1509912.0\n",
      "  Epoch 14 of 15.Change in loss after epoch 14: 1466700.0\n",
      "  Epoch 15 of 15.Change in loss after epoch 15: 1430540.0\n",
      "Saving model with name: C:\\Users\\joris\\Documents\\eScience_data\\data\\trained_models\\spec2vec_librarymatchingnew3.model\n"
     ]
    }
   ],
   "source": [
    "#train new model on allpositive without new selected spectra\n",
    "\n",
    "#path_models = os.path.join(path_data, \"trained_models\")\n",
    "#new3_model_file = os.path.join(path_models, \"spec2vec_librarymatchingnew3.model\")\n",
    "\n",
    "#iterations = [15]\n",
    "\n",
    "# Train model with default parameters\n",
    "#new3_model = train_new_word2vec_model(new3_documents_library_s2v, iterations, new3_model_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  Epoch 1 of 15.Change in loss after epoch 1: 5658305.5\n",
      "  Epoch 2 of 15.Change in loss after epoch 2: 4313616.5\n",
      "  Epoch 3 of 15.Change in loss after epoch 3: 3782774.0\n",
      "  Epoch 4 of 15.Change in loss after epoch 4: 3483726.0\n",
      "  Epoch 5 of 15.Change in loss after epoch 5: 2885978.0\n",
      "  Epoch 6 of 15.Change in loss after epoch 6: 2798772.0\n",
      "  Epoch 7 of 15.Change in loss after epoch 7: 2649280.0\n",
      "  Epoch 8 of 15.Change in loss after epoch 8: 2627094.0\n",
      "  Epoch 9 of 15.Change in loss after epoch 9: 2497248.0\n",
      "  Epoch 10 of 15.Change in loss after epoch 10: 2407102.0\n",
      "  Epoch 11 of 15.Change in loss after epoch 11: 1683920.0\n",
      "  Epoch 12 of 15.Change in loss after epoch 12: 1479412.0\n",
      "  Epoch 13 of 15.Change in loss after epoch 13: 1437412.0\n",
      "  Epoch 14 of 15.Change in loss after epoch 14: 1446216.0\n",
      "  Epoch 15 of 15.Change in loss after epoch 15: 1420592.0\n",
      "Saving model with name: C:\\Users\\joris\\Documents\\eScience_data\\data\\trained_models\\spec2vec_librarymatching_plus_unique.model\n"
     ]
    }
   ],
   "source": [
    "#train new model on allpositive without 1000 spectra >5 occurrence and 1000 unique spectra\n",
    "\n",
    "#path_models = os.path.join(path_data, \"trained_models\")\n",
    "#old_and_unique_model_file = os.path.join(path_models, \"spec2vec_librarymatching_plus_unique.model\")\n",
    "\n",
    "#iterations = [15]\n",
    "\n",
    "# Train model with default parameters\n",
    "#old_and_unique_model = train_new_word2vec_model(old_and_unique_documents_library_s2v, iterations, old_and_unique_model_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  Epoch 1 of 15.Change in loss after epoch 1: 5690089.0\n",
      "  Epoch 2 of 15.Change in loss after epoch 2: 4205896.0\n",
      "  Epoch 3 of 15.Change in loss after epoch 3: 3737722.0\n",
      "  Epoch 4 of 15.Change in loss after epoch 4: 3461079.0\n",
      "  Epoch 5 of 15.Change in loss after epoch 5: 2867512.0\n",
      "  Epoch 6 of 15.Change in loss after epoch 6: 2762174.0\n",
      "  Epoch 7 of 15.Change in loss after epoch 7: 2642014.0\n",
      "  Epoch 8 of 15.Change in loss after epoch 8: 2542064.0\n",
      "  Epoch 9 of 15.Change in loss after epoch 9: 2485120.0\n",
      "  Epoch 10 of 15.Change in loss after epoch 10: 2422124.0\n",
      "  Epoch 11 of 15.Change in loss after epoch 11: 1823130.0\n",
      "  Epoch 12 of 15.Change in loss after epoch 12: 1482092.0\n",
      "  Epoch 13 of 15.Change in loss after epoch 13: 1452004.0\n",
      "  Epoch 14 of 15.Change in loss after epoch 14: 1402236.0\n",
      "  Epoch 15 of 15.Change in loss after epoch 15: 1394148.0\n",
      "Saving model with name: C:\\Users\\joris\\Documents\\eScience_data\\data\\trained_models\\spec2vec_librarymatchingnew_plus_unique2.model\n"
     ]
    }
   ],
   "source": [
    "#train new model on allpositive without 1000 spectra >5 occurrence and 1000 unique spectra\n",
    "\n",
    "path_models = os.path.join(path_data, \"trained_models\")\n",
    "new_and_unique2_model_file = os.path.join(path_models, \"spec2vec_librarymatchingnew_plus_unique2.model\")\n",
    "\n",
    "iterations = [15]\n",
    "\n",
    "# Train model with default parameters\n",
    "new_and_unique2_model = train_new_word2vec_model(new_and_unique2_documents_library_s2v, iterations, new_and_unique2_model_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Actual library matching"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "sys.path.insert(0, \"C:\\\\Users\\\\joris\\\\Documents\\\\eScience_data\\\\spec2vec_gnps_data_analysis\\\\custom_functions\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "from library_search import library_matching"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Load word2vec model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "path_models = os.path.join(path_data, \"trained_models\")\n",
    "model_file = os.path.join(path_models, \"spec2vec_librarymatching_iter_15.model\")\n",
    "model = gensim.models.Word2Vec.load(model_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Word2Vec(vocab=115910, size=300, alpha=0.025)\n"
     ]
    }
   ],
   "source": [
    "print(model)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "documents_query_s2v = [SpectrumDocument(spectrums_s2v[i], n_decimals=2) for i in selected_spectra]\n",
    "outfile = os.path.join(path_data, 'documents_query_s2v.pickle')\n",
    "print(outfile)\n",
    "if not os.path.exists(outfile):\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(documents_query_s2v, outf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1000"
      ]
     },
     "execution_count": 96,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(documents_query_s2v)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'SRRQPVVYXBTRQK-XMMPIXPASA-N'"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "documents_query_s2v[0]._obj.get(\"inchikey\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 218,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C:\\Users\\joris\\Documents\\eScience_data\\data\\new_documents_query_s2v.pickle\n"
     ]
    }
   ],
   "source": [
    "#new model\n",
    "outfile = os.path.join(path_data, 'new_documents_query_s2v.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        new_documents_query_s2v = pickle.load(inf)\n",
    "else:\n",
    "    new_documents_query_s2v = [SpectrumDocument(spectrums_s2v[i], n_decimals=2) for i in new_selected_spectra]\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(new_documents_query_s2v, outf)\n",
    "\n",
    "path_models = os.path.join(path_data, \"trained_models\")\n",
    "new_model_file = os.path.join(path_models, \"spec2vec_librarymatchingnew_iter_15.model\")\n",
    "new_model = gensim.models.Word2Vec.load(model_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [],
   "source": [
    "#new models\n",
    "new2_documents_query_s2v = [SpectrumDocument(spectrums_s2v[i], n_decimals=2) for i in new2_selected_spectra]\n",
    "path_models = os.path.join(path_data, \"trained_models\")\n",
    "new2_model_file = os.path.join(path_models, \"spec2vec_librarymatchingnew2.model\")\n",
    "new2_model = gensim.models.Word2Vec.load(new2_model_file)\n",
    "\n",
    "#number 3\n",
    "new3_documents_query_s2v = [SpectrumDocument(spectrums_s2v[i], n_decimals=2) for i in new3_selected_spectra]\n",
    "path_models = os.path.join(path_data, \"trained_models\")\n",
    "new3_model_file = os.path.join(path_models, \"spec2vec_librarymatchingnew3.model\")\n",
    "new3_model = gensim.models.Word2Vec.load(new3_model_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C:\\Users\\joris\\Documents\\eScience_data\\data\\old_and_unique_documents_query_s2v.pickle\n"
     ]
    }
   ],
   "source": [
    "#with unique\n",
    "outfile = os.path.join(path_data, 'old_and_unique_documents_query_s2v.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        old_and_unique_documents_query_s2v = pickle.load(inf)\n",
    "else:\n",
    "    old_and_unique_documents_query_s2v = [SpectrumDocument(spectrums_s2v[i], n_decimals=2) for i in old_and_unique_test_set]\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(old_and_unique_documents_query_s2v, outf)\n",
    "\n",
    "path_models = os.path.join(path_data, \"trained_models\")\n",
    "old_and_unique_model_file = os.path.join(path_models, \"spec2vec_librarymatching_plus_unique.model\")\n",
    "old_and_unique_model = gensim.models.Word2Vec.load(old_and_unique_model_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C:\\Users\\joris\\Documents\\eScience_data\\data\\new_and_unique2_documents_query_s2v.pickle\n"
     ]
    }
   ],
   "source": [
    "#new with unique2\n",
    "outfile = os.path.join(path_data, 'new_and_unique2_documents_query_s2v.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        new_and_unique2_documents_query_s2v = pickle.load(inf)\n",
    "else:\n",
    "    new_and_unique2_documents_query_s2v = [SpectrumDocument(spectrums_s2v[i], n_decimals=2) for i in new_and_unique2_test_set]\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(new_and_unique2_documents_query_s2v, outf)\n",
    "\n",
    "path_models = os.path.join(path_data, \"trained_models\")\n",
    "new_and_unique2_model_file = os.path.join(path_models, \"spec2vec_librarymatchingnew_plus_unique2.model\")\n",
    "new_and_unique2_model = gensim.models.Word2Vec.load(new_and_unique2_model_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Correct a few (< 10) spectrums with missing precursor_mz (in place)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2765\n",
      "parent mass new: 547.2247235480093 former was: -1.00727645199076\n",
      "precursor_mz new: 548.232 former was: 0.0\n",
      "2766\n",
      "parent mass new: 589.2357235480093 former was: -1.00727645199076\n",
      "precursor_mz new: 590.243 former was: 0.0\n",
      "3942\n",
      "parent mass new: 466.2697235480092 former was: -1.00727645199076\n",
      "precursor_mz new: 467.277 former was: 0.0\n",
      "3974\n",
      "parent mass new: 674.4887235480093 former was: -1.00727645199076\n",
      "precursor_mz new: 675.496 former was: 0.0\n",
      "3975\n",
      "parent mass new: 691.5147235480093 former was: -1.00727645199076\n",
      "precursor_mz new: 692.522 former was: 0.0\n",
      "3976\n",
      "parent mass new: 696.4707235480092 former was: -1.00727645199076\n",
      "precursor_mz new: 697.478 former was: 0.0\n",
      "3983\n",
      "parent mass new: 629.6107235480093 former was: -1.00727645199076\n",
      "precursor_mz new: 630.618 former was: 0.0\n",
      "3984\n",
      "parent mass new: 491.47072354800923 former was: -1.00727645199076\n",
      "precursor_mz new: 492.478 former was: 0.0\n",
      "3985\n",
      "parent mass new: 531.4627235480093 former was: -1.00727645199076\n",
      "precursor_mz new: 532.47 former was: 0.0\n",
      "21107\n",
      "parent mass new: 199.09972354800925 former was: -1.00727645199076\n",
      "precursor_mz new: 200.107 former was: 0.0\n",
      "21364\n",
      "parent mass new: 517.3137235480093 former was: -1.00727645199076\n",
      "precursor_mz new: 518.321 former was: 0.0\n",
      "22053\n",
      "parent mass new: 208.18272354800925 former was: -1.00727645199076\n",
      "precursor_mz new: 209.19 former was: 0.0\n"
     ]
    }
   ],
   "source": [
    "for i, spec in enumerate(spectrums_s2v):\n",
    "    if not spec.get(\"precursor_mz\"):\n",
    "        print(i)\n",
    "        fix_mass = 0.0\n",
    "        for history in spec.metadata['annotation_history']:\n",
    "            fix_mass_test = float(history['Precursor_MZ'])\n",
    "            fix_mass = max(fix_mass, fix_mass_test)\n",
    "        charge = spec.get(\"charge\")\n",
    "        protons_mass = 1.00727645199076 * charge\n",
    "        precursor_mass = fix_mass * abs(charge)\n",
    "        parent_mass = precursor_mass - protons_mass\n",
    "        if parent_mass:\n",
    "            print(\"parent mass new:\", parent_mass, \"former was:\", spec.get(\"parent_mass\"))\n",
    "            spec.set(\"parent_mass\", parent_mass)\n",
    "        print(\"precursor_mz new:\", fix_mass, \"former was:\", spec.get(\"precursor_mz\"))\n",
    "        spec.set(\"precursor_mz\", fix_mass)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Run Spec2Vec and parentmass based library matching for 1000 against 94320 spectra"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Pre-selection includes spec2vec top 20.\n"
     ]
    }
   ],
   "source": [
    "import pickle\n",
    "outfile = os.path.join(path_data, 'found_matches_s2v.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        found_matches_s2v = pickle.load(inf)\n",
    "else:\n",
    "    found_matches_s2v = library_matching(documents_query_s2v, documents_library_s2v,\n",
    "                                     model,\n",
    "                                     presearch_based_on=[\"parentmass\", \"spec2vec-top20\"],\n",
    "                                     ignore_non_annotated=True,\n",
    "                                     intensity_weighting_power=0.5,\n",
    "                                     allowed_missing_percentage=5.0,\n",
    "                                     cosine_tol=0.005,\n",
    "                                     mass_tolerance=1.0)\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(found_matches_s2v, outf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Pre-selection includes spec2vec top 20.\n"
     ]
    }
   ],
   "source": [
    "new_found_matches_s2v = library_matching(new_documents_query_s2v, new_documents_library_s2v,\n",
    "                                     new_model,\n",
    "                                     presearch_based_on=[\"parentmass\", \"spec2vec-top20\"],\n",
    "                                     ignore_non_annotated=True,\n",
    "                                     intensity_weighting_power=0.5,\n",
    "                                     allowed_missing_percentage=5.0,\n",
    "                                     cosine_tol=0.005,\n",
    "                                     mass_tolerance=1.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Pre-selection includes spec2vec top 20.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.21%.\n",
      "Found 2 word(s) missing in the model. Weighted fraction not covered is 0.38%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.44%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.24%.\n",
      "Found 2 word(s) missing in the model. Weighted fraction not covered is 0.41%.\n",
      "Found 7 word(s) missing in the model. Weighted fraction not covered is 1.19%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.16%.\n",
      "Found 11 word(s) missing in the model. Weighted fraction not covered is 1.79%.\n",
      "Pre-selection includes spec2vec top 20.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.51%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.18%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.16%.\n",
      "Found 5 word(s) missing in the model. Weighted fraction not covered is 1.25%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.33%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.14%.\n",
      "Found 3 word(s) missing in the model. Weighted fraction not covered is 0.98%.\n",
      "Found 21 word(s) missing in the model. Weighted fraction not covered is 4.97%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.27%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 4.32%.\n",
      "Pre-selection includes spec2vec top 20.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.18%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.24%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.09%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.16%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.21%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.28%.\n",
      "Found 2 word(s) missing in the model. Weighted fraction not covered is 0.56%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.16%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.29%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.18%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.39%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.18%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.25%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.48%.\n"
     ]
    }
   ],
   "source": [
    "new2_found_matches_s2v = library_matching(new2_documents_query_s2v, new2_documents_library_s2v,\n",
    "                                     new2_model,\n",
    "                                     presearch_based_on=[\"parentmass\", \"spec2vec-top20\"],\n",
    "                                     ignore_non_annotated=True,\n",
    "                                     intensity_weighting_power=0.5,\n",
    "                                     allowed_missing_percentage=5.0,\n",
    "                                     cosine_tol=0.005,\n",
    "                                     mass_tolerance=1.0)\n",
    "new3_found_matches_s2v = library_matching(new3_documents_query_s2v, new3_documents_library_s2v,\n",
    "                                     new3_model,\n",
    "                                     presearch_based_on=[\"parentmass\", \"spec2vec-top20\"],\n",
    "                                     ignore_non_annotated=True,\n",
    "                                     intensity_weighting_power=0.5,\n",
    "                                     allowed_missing_percentage=5.0,\n",
    "                                     cosine_tol=0.005,\n",
    "                                     mass_tolerance=1.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Pre-selection includes spec2vec top 20.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.18%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.24%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.09%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.16%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.21%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.28%.\n",
      "Found 2 word(s) missing in the model. Weighted fraction not covered is 0.56%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.16%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.29%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.18%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.39%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.18%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.25%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.48%.\n"
     ]
    }
   ],
   "source": [
    "import pickle\n",
    "outfile = os.path.join(path_data, 'old_and_unique_found_matches_s2v.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        old_and_unique_found_matches_s2v = pickle.load(inf)\n",
    "else:\n",
    "    old_and_unique_found_matches_s2v = library_matching(old_and_unique_documents_query_s2v, old_and_unique_documents_library_s2v,\n",
    "                                     old_and_unique_model,\n",
    "                                     presearch_based_on=[\"parentmass\", \"spec2vec-top20\"],\n",
    "                                     ignore_non_annotated=True,\n",
    "                                     intensity_weighting_power=0.5,\n",
    "                                     allowed_missing_percentage=5.0,\n",
    "                                     cosine_tol=0.005,\n",
    "                                     mass_tolerance=1.0)\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(old_and_unique_found_matches_s2v, outf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C:\\Users\\joris\\Documents\\eScience_data\\data\\new_and_unique2_found_matches_s2v.pickle\n",
      "Pre-selection includes spec2vec top 20.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.26%.\n",
      "Found 2 word(s) missing in the model. Weighted fraction not covered is 6.32%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.30%.\n",
      "Found 2 word(s) missing in the model. Weighted fraction not covered is 0.29%.\n",
      "Found 3 word(s) missing in the model. Weighted fraction not covered is 0.44%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 1.29%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.07%.\n",
      "Found 2 word(s) missing in the model. Weighted fraction not covered is 8.16%.\n",
      "Found 3 word(s) missing in the model. Weighted fraction not covered is 21.90%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.17%.\n",
      "Found 1 word(s) missing in the model. Weighted fraction not covered is 0.51%.\n"
     ]
    }
   ],
   "source": [
    "import pickle\n",
    "outfile = os.path.join(path_data, 'new_and_unique2_found_matches_s2v.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        new_and_unique2_found_matches_s2v = pickle.load(inf)\n",
    "else:\n",
    "    new_and_unique2_found_matches_s2v = library_matching(new_and_unique2_documents_query_s2v, new_and_unique2_documents_library_s2v,\n",
    "                                     new_and_unique2_model,\n",
    "                                     presearch_based_on=[\"parentmass\", \"spec2vec-top20\"],\n",
    "                                     ignore_non_annotated=True,\n",
    "                                     intensity_weighting_power=0.5,\n",
    "                                     allowed_missing_percentage=30.0,\n",
    "                                     cosine_tol=0.005,\n",
    "                                     mass_tolerance=1.0)\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(new_and_unique2_found_matches_s2v, outf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>46341</th>\n",
       "      <td>0.792853</td>\n",
       "      <td>3</td>\n",
       "      <td>0.792853</td>\n",
       "      <td>3</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.838697</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52743</th>\n",
       "      <td>0.766401</td>\n",
       "      <td>1</td>\n",
       "      <td>0.766401</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.589020</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34539</th>\n",
       "      <td>0.241939</td>\n",
       "      <td>1</td>\n",
       "      <td>0.955195</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.585604</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47562</th>\n",
       "      <td>0.010289</td>\n",
       "      <td>1</td>\n",
       "      <td>0.010289</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.584888</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35469</th>\n",
       "      <td>0.243206</td>\n",
       "      <td>1</td>\n",
       "      <td>0.243206</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.565561</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14676</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.012161</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.049336</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3754</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.049613</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>69274</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.052223</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>81463</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.055963</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>81464</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.070338</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>237 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "46341      0.792853               3          0.792853                   3   \n",
       "52743      0.766401               1          0.766401                   1   \n",
       "34539      0.241939               1          0.955195                   2   \n",
       "47562      0.010289               1          0.010289                   1   \n",
       "35469      0.243206               1          0.243206                   1   \n",
       "...             ...             ...               ...                 ...   \n",
       "14676      0.000000               0          0.012161                   1   \n",
       "3754       0.000000               0          0.000000                   0   \n",
       "69274      0.000000               0          0.000000                   0   \n",
       "81463      0.000000               0          0.000000                   0   \n",
       "81464      0.000000               0          0.000000                   0   \n",
       "\n",
       "       mass_match  s2v_score  \n",
       "46341         1.0   0.838697  \n",
       "52743         0.0   0.589020  \n",
       "34539         0.0   0.585604  \n",
       "47562         1.0   0.584888  \n",
       "35469         0.0   0.565561  \n",
       "...           ...        ...  \n",
       "14676         1.0  -0.049336  \n",
       "3754          1.0  -0.049613  \n",
       "69274         1.0  -0.052223  \n",
       "81463         1.0  -0.055963  \n",
       "81464         1.0  -0.070338  \n",
       "\n",
       "[237 rows x 6 columns]"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "new_found_matches_s2v[0].sort_values(by=['s2v_score'],ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>46341</th>\n",
       "      <td>0.792853</td>\n",
       "      <td>3</td>\n",
       "      <td>0.792853</td>\n",
       "      <td>3</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.838697</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52743</th>\n",
       "      <td>0.766401</td>\n",
       "      <td>1</td>\n",
       "      <td>0.766401</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.589020</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34539</th>\n",
       "      <td>0.241939</td>\n",
       "      <td>1</td>\n",
       "      <td>0.955195</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.585604</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47562</th>\n",
       "      <td>0.010289</td>\n",
       "      <td>1</td>\n",
       "      <td>0.010289</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.584888</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35469</th>\n",
       "      <td>0.243206</td>\n",
       "      <td>1</td>\n",
       "      <td>0.243206</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.565561</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14676</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.012161</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.049336</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3754</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.049613</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>69274</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.052223</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>81463</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.055963</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>81464</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.070338</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>237 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "46341      0.792853               3          0.792853                   3   \n",
       "52743      0.766401               1          0.766401                   1   \n",
       "34539      0.241939               1          0.955195                   2   \n",
       "47562      0.010289               1          0.010289                   1   \n",
       "35469      0.243206               1          0.243206                   1   \n",
       "...             ...             ...               ...                 ...   \n",
       "14676      0.000000               0          0.012161                   1   \n",
       "3754       0.000000               0          0.000000                   0   \n",
       "69274      0.000000               0          0.000000                   0   \n",
       "81463      0.000000               0          0.000000                   0   \n",
       "81464      0.000000               0          0.000000                   0   \n",
       "\n",
       "       mass_match  s2v_score  \n",
       "46341         1.0   0.838697  \n",
       "52743         0.0   0.589020  \n",
       "34539         0.0   0.585604  \n",
       "47562         1.0   0.584888  \n",
       "35469         0.0   0.565561  \n",
       "...           ...        ...  \n",
       "14676         1.0  -0.049336  \n",
       "3754          1.0  -0.049613  \n",
       "69274         1.0  -0.052223  \n",
       "81463         1.0  -0.055963  \n",
       "81464         1.0  -0.070338  \n",
       "\n",
       "[237 rows x 6 columns]"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "new_found_matches_s2v[0].sort_values(by=['s2v_score'],ascending=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### How do the old and new test set compare?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(20000, 20000, 239233, 241054)"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from itertools import chain\n",
    "old20_all_s2v_scores_list = [sorted(found_matches_s2v[i]['s2v_score'],reverse=True)[:20] for i in range(len(found_matches_s2v))]\n",
    "old20_all_s2v_scores = list(chain.from_iterable(old20_all_s2v_scores_list))\n",
    "new20_all_s2v_scores_list = [sorted(new_found_matches_s2v[i]['s2v_score'],reverse=True)[:20] for i in range(len(new_found_matches_s2v))]\n",
    "new20_all_s2v_scores = list(chain.from_iterable(new20_all_s2v_scores_list))\n",
    "\n",
    "old_all_s2v_scores_list = [found_matches_s2v[i]['s2v_score'] for i in range(len(found_matches_s2v))]\n",
    "old_all_s2v_scores = list(chain.from_iterable(old_all_s2v_scores_list))\n",
    "new_all_s2v_scores_list = [new_found_matches_s2v[i]['s2v_score'] for i in range(len(new_found_matches_s2v))]\n",
    "new_all_s2v_scores = list(chain.from_iterable(new_all_s2v_scores_list))\n",
    "\n",
    "len(old20_all_s2v_scores), len(new20_all_s2v_scores), len(old_all_s2v_scores), len(new_all_s2v_scores)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(12,6))\n",
    "bns = np.linspace(0,1,100)\n",
    "plt.hist(old20_all_s2v_scores, bins = bns, alpha=0.5, label='old_top20_s2v_scores')\n",
    "plt.hist(new20_all_s2v_scores, bins = bns, alpha=0.5, label='new_top20_s2v_scores')\n",
    "plt.legend(loc='upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(12,6))\n",
    "bns = np.linspace(0,1,100)\n",
    "plt.hist(old_all_s2v_scores, bins = bns, alpha=0.5, label='old_s2v_scores')\n",
    "plt.hist(new_all_s2v_scores, bins = bns, alpha=0.5, label='new_s2v_scores')\n",
    "plt.legend(loc='upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.boxplot([old20_all_s2v_scores, new20_all_s2v_scores])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>old20</th>\n",
       "      <th>new20</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>20000.000000</td>\n",
       "      <td>20000.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>0.648347</td>\n",
       "      <td>0.650032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.180152</td>\n",
       "      <td>0.180522</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.199394</td>\n",
       "      <td>0.206990</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.522621</td>\n",
       "      <td>0.524059</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.647679</td>\n",
       "      <td>0.648011</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>0.774202</td>\n",
       "      <td>0.778094</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              old20         new20\n",
       "count  20000.000000  20000.000000\n",
       "mean       0.648347      0.650032\n",
       "std        0.180152      0.180522\n",
       "min        0.199394      0.206990\n",
       "25%        0.522621      0.524059\n",
       "50%        0.647679      0.648011\n",
       "75%        0.774202      0.778094\n",
       "max        1.000000      1.000000"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "compare_old_new20 = pd.DataFrame({'old20':old20_all_s2v_scores, 'new20':new20_all_s2v_scores})\n",
    "compare_old_new20.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "old_parent_masses = [documents_query_s2v[i]._obj.get('parent_mass') for i in range(len(documents_query_s2v))]\n",
    "new_parent_masses = [new_documents_query_s2v[i]._obj.get('parent_mass') for i in range(len(new_documents_query_s2v))]\n",
    "plt.figure(figsize=(12,6))\n",
    "bns = np.linspace(0,max([mass for mass in (old_parent_masses+new_parent_masses)]), 130)\n",
    "plt.hist(old_parent_masses, bins=bns, alpha=0.5, label='old_parent_masses')\n",
    "plt.hist(new_parent_masses, bins=bns, alpha=0.5, label='new_parent_masses')\n",
    "plt.legend(loc='upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.boxplot([old_parent_masses,new_parent_masses])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>old_parent_masses</th>\n",
       "      <th>new_parent_masses</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>1000.000000</td>\n",
       "      <td>1000.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>453.683394</td>\n",
       "      <td>455.946632</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>245.729585</td>\n",
       "      <td>261.888464</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>44.992724</td>\n",
       "      <td>73.992724</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>283.125974</td>\n",
       "      <td>280.148724</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>387.216224</td>\n",
       "      <td>372.228224</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>573.291974</td>\n",
       "      <td>563.890474</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1756.982724</td>\n",
       "      <td>1892.562724</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       old_parent_masses  new_parent_masses\n",
       "count        1000.000000        1000.000000\n",
       "mean          453.683394         455.946632\n",
       "std           245.729585         261.888464\n",
       "min            44.992724          73.992724\n",
       "25%           283.125974         280.148724\n",
       "50%           387.216224         372.228224\n",
       "75%           573.291974         563.890474\n",
       "max          1756.982724        1892.562724"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame({'old_parent_masses':old_parent_masses, 'new_parent_masses':new_parent_masses}).describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 192,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "([3, 0, 0, 0, 0, 0, 1, 11, 2, 2], 886)"
      ]
     },
     "execution_count": 192,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "new_best_match_position = []\n",
    "topn = 20\n",
    "for match in new_found_matches_s2v_match_perc_pmass_sim:\n",
    "    match = match.sort_values(by=['s2v_score'],ascending=False).iloc[:topn]\n",
    "    for i in range(match.shape[0]):\n",
    "        if match.iloc[i]['label'] == 1:\n",
    "            new_best_match_position.append(i)\n",
    "            break\n",
    "new_best_match_position[:10], len(new_best_match_position)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "([0, 0, 0, 0, 0, 0, 0, 0, 1, 2], 898)"
      ]
     },
     "execution_count": 193,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_match_position = []\n",
    "for match in found_matches_s2v_match_perc_pmass_sim:\n",
    "    match = match.sort_values(by=['s2v_score'],ascending=False).iloc[:topn]\n",
    "    for i in range(match.shape[0]):\n",
    "        if match.iloc[i]['label'] == 1:\n",
    "            best_match_position.append(i)\n",
    "            break\n",
    "best_match_position[:10], len(best_match_position)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "720"
      ]
     },
     "execution_count": 200,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_match_position.count(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                0\n",
      "count  898.000000\n",
      "mean     0.871938\n",
      "std      2.672848\n",
      "min      0.000000\n",
      "25%      0.000000\n",
      "50%      0.000000\n",
      "75%      0.000000\n",
      "max     19.000000\n",
      "                0\n",
      "count  886.000000\n",
      "mean     0.863431\n",
      "std      2.541959\n",
      "min      0.000000\n",
      "25%      0.000000\n",
      "50%      0.000000\n",
      "75%      0.000000\n",
      "max     17.000000\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(pd.DataFrame(best_match_position).describe())\n",
    "print(pd.DataFrame(new_best_match_position).describe())\n",
    "plt.figure(figsize=(12,6))\n",
    "bns = np.linspace(0, 20, 20)\n",
    "plt.hist(best_match_position, bins=bns, alpha=0.5, label='old_best_match_position')\n",
    "plt.hist(new_best_match_position, bins=bns, alpha=0.5, label='new_best_match_position')\n",
    "plt.legend(loc='upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "metadata": {},
   "outputs": [],
   "source": [
    "#the number of found true matches per query\n",
    "new_num_matches = [match[match['label']==1].shape[0] for match in new_found_matches_s2v_match_perc_pmass_sim]\n",
    "old_num_matches = [match[match['label']==1].shape[0] for match in found_matches_s2v_match_perc_pmass_sim]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>new_fnum_matches_per_query</th>\n",
       "      <th>old_fnum_matches_per_query</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>1000.000000</td>\n",
       "      <td>1000.00000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>9.151000</td>\n",
       "      <td>10.62700</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>19.840964</td>\n",
       "      <td>24.36213</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>3.000000</td>\n",
       "      <td>3.00000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>5.000000</td>\n",
       "      <td>5.00000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>9.000000</td>\n",
       "      <td>9.00000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>425.000000</td>\n",
       "      <td>431.00000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       new_fnum_matches_per_query  old_fnum_matches_per_query\n",
       "count                 1000.000000                  1000.00000\n",
       "mean                     9.151000                    10.62700\n",
       "std                     19.840964                    24.36213\n",
       "min                      0.000000                     0.00000\n",
       "25%                      3.000000                     3.00000\n",
       "50%                      5.000000                     5.00000\n",
       "75%                      9.000000                     9.00000\n",
       "max                    425.000000                   431.00000"
      ]
     },
     "execution_count": 201,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame({'new_fnum_matches_per_query':new_num_matches, 'old_fnum_matches_per_query':old_num_matches}).describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 202,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(12,6))\n",
    "bns = np.linspace(0, 100, 100)\n",
    "plt.hist(old_num_matches, bins=bns, alpha=0.5, label='old_fnum_matches_per_query')\n",
    "plt.hist(new_num_matches, bins=bns, alpha=0.5, label='new_fnum_matches_per_query')\n",
    "plt.legend(loc='upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "metadata": {},
   "outputs": [
    {
     "ename": "KeyError",
     "evalue": "'occurences'",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[1;32mc:\\users\\joris\\documents\\miniconda3\\envs\\spec2vec_analysis\\lib\\site-packages\\pandas\\core\\indexes\\base.py\u001b[0m in \u001b[0;36mget_loc\u001b[1;34m(self, key, method, tolerance)\u001b[0m\n\u001b[0;32m   2645\u001b[0m             \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2646\u001b[1;33m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2647\u001b[0m             \u001b[1;32mexcept\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\index_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.index.Int64Engine._check_type\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;31mKeyError\u001b[0m: 'occurences'",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[1;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-204-71ea56411faf>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m      1\u001b[0m \u001b[1;31m#the number of matching inchikey per query\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      2\u001b[0m \u001b[0mnew_all_true_matches\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnew_suitable_inchikeys\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'occurences'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mnew_selection\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m-\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 3\u001b[1;33m \u001b[0mold_all_true_matches\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0msuitable_inchikeys\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'occurences'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mselection\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m-\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      4\u001b[0m \u001b[0mpd\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mDataFrame\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m{\u001b[0m\u001b[1;34m'old_all_true_matches'\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mold_all_true_matches\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;34m'new_all_true_matches'\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mnew_all_true_matches\u001b[0m\u001b[1;33m}\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdescribe\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mc:\\users\\joris\\documents\\miniconda3\\envs\\spec2vec_analysis\\lib\\site-packages\\pandas\\core\\frame.py\u001b[0m in \u001b[0;36m__getitem__\u001b[1;34m(self, key)\u001b[0m\n\u001b[0;32m   2798\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mnlevels\u001b[0m \u001b[1;33m>\u001b[0m \u001b[1;36m1\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2799\u001b[0m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_getitem_multilevel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2800\u001b[1;33m             \u001b[0mindexer\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2801\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mis_integer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mindexer\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2802\u001b[0m                 \u001b[0mindexer\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;33m[\u001b[0m\u001b[0mindexer\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mc:\\users\\joris\\documents\\miniconda3\\envs\\spec2vec_analysis\\lib\\site-packages\\pandas\\core\\indexes\\range.py\u001b[0m in \u001b[0;36mget_loc\u001b[1;34m(self, key, method, tolerance)\u001b[0m\n\u001b[0;32m    351\u001b[0m             \u001b[1;32mexcept\u001b[0m \u001b[0mValueError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    352\u001b[0m                 \u001b[1;32mraise\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 353\u001b[1;33m         \u001b[1;32mreturn\u001b[0m \u001b[0msuper\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmethod\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mmethod\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtolerance\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mtolerance\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    354\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    355\u001b[0m     \u001b[1;33m@\u001b[0m\u001b[0mAppender\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0m_index_shared_docs\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m\"get_indexer\"\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mc:\\users\\joris\\documents\\miniconda3\\envs\\spec2vec_analysis\\lib\\site-packages\\pandas\\core\\indexes\\base.py\u001b[0m in \u001b[0;36mget_loc\u001b[1;34m(self, key, method, tolerance)\u001b[0m\n\u001b[0;32m   2646\u001b[0m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2647\u001b[0m             \u001b[1;32mexcept\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2648\u001b[1;33m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_maybe_cast_indexer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2649\u001b[0m         \u001b[0mindexer\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_indexer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmethod\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mmethod\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtolerance\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mtolerance\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2650\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mindexer\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mndim\u001b[0m \u001b[1;33m>\u001b[0m \u001b[1;36m1\u001b[0m \u001b[1;32mor\u001b[0m \u001b[0mindexer\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msize\u001b[0m \u001b[1;33m>\u001b[0m \u001b[1;36m1\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\index_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.index.Int64Engine._check_type\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;31mKeyError\u001b[0m: 'occurences'"
     ]
    }
   ],
   "source": [
    "#the number of matching inchikey per query\n",
    "new_all_true_matches = new_suitable_inchikeys['occurences'].values[new_selection]-1\n",
    "old_all_true_matches = suitable_inchikeys['occurences'].values[selection]-1\n",
    "pd.DataFrame({'old_all_true_matches':old_all_true_matches, 'new_all_true_matches':new_all_true_matches}).describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.figure(figsize=(18,6))\n",
    "bns = np.linspace(0, 200, 200)\n",
    "plt.hist(old_all_true_matches, bins=bns, alpha=0.5, label='old_all_true_matches')\n",
    "plt.hist(new_all_true_matches, bins=bns, alpha=0.5, label='new_all_true_matches')\n",
    "plt.hist(suitable_inchikeys['occurences']-1, bins=bns, alpha=0.4, label='all_true_matches')\n",
    "plt.legend(loc='upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#do the best matches have mass match?\n",
    "new_best_match_have_mm = []\n",
    "for match in new_found_matches_s2v_match_perc_pmass_sim:\n",
    "    match = match.sort_values(by=['s2v_score'],ascending=False)\n",
    "    for i in range(match.shape[0]):\n",
    "        if match.iloc[i]['label'] == 1:\n",
    "            res = 0\n",
    "            if match.iloc[i]['mass_match'] == 1:\n",
    "                res = 1\n",
    "            new_best_match_have_mm.append(res)\n",
    "            break\n",
    "print('Number of new best matches that has mass match:', new_best_match_have_mm.count(1))\n",
    "print('Total of new best matches:', len(new_best_match_have_mm))\n",
    "\n",
    "old_best_match_have_mm = []\n",
    "for match in found_matches_s2v_match_perc_pmass_sim:\n",
    "    match = match.sort_values(by=['s2v_score'],ascending=False)\n",
    "    for i in range(match.shape[0]):\n",
    "        if match.iloc[i]['label'] == 1:\n",
    "            res = 0\n",
    "            if match.iloc[i]['mass_match'] == 1:\n",
    "                res = 1\n",
    "            old_best_match_have_mm.append(res)\n",
    "            break\n",
    "print('Number of old best matches that has mass match:', old_best_match_have_mm.count(1))\n",
    "print('Total of old best matches:', len(old_best_match_have_mm))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Testing performance and training random forest models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "46003\n",
      "SRRQPVVYXBTRQK-XMMPIXPASA-N\n",
      "SRRQPVVYXBTRQK-XMMPIXPASA-N\n"
     ]
    }
   ],
   "source": [
    "threshold = 0.5\n",
    "min_match = 1\n",
    "ID = 0\n",
    "df_select = found_matches_s2v[ID][(found_matches_s2v[ID]['mass_match'] == 1)\n",
    "                     & (found_matches_s2v[ID]['cosine_score'] > threshold)\n",
    "                     & (found_matches_s2v[ID]['cosine_matches'] >= min_match)]\n",
    "best_match_ID = df_select.sort_values(by=['cosine_score'], ascending=False).index[0]\n",
    "print(best_match_ID)\n",
    "print(documents_library_s2v[best_match_ID]._obj.get(\"inchikey\"))\n",
    "print(documents_query_s2v[ID]._obj.get(\"inchikey\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Checking matches for S2V score > 0.0\n",
      "Checking matches for S2V score > 0.05\n",
      "Checking matches for S2V score > 0.1\n",
      "Checking matches for S2V score > 0.15000000000000002\n",
      "Checking matches for S2V score > 0.2\n",
      "Checking matches for S2V score > 0.25\n",
      "Checking matches for S2V score > 0.30000000000000004\n",
      "Checking matches for S2V score > 0.35000000000000003\n",
      "Checking matches for S2V score > 0.4\n",
      "Checking matches for S2V score > 0.45\n",
      "Checking matches for S2V score > 0.5\n",
      "Checking matches for S2V score > 0.55\n",
      "Checking matches for S2V score > 0.6000000000000001\n",
      "Checking matches for S2V score > 0.65\n",
      "Checking matches for S2V score > 0.7000000000000001\n",
      "Checking matches for S2V score > 0.75\n",
      "Checking matches for S2V score > 0.8\n",
      "Checking matches for S2V score > 0.8500000000000001\n",
      "Checking matches for S2V score > 0.9\n",
      "Checking matches for S2V score > 0.9500000000000001\n"
     ]
    }
   ],
   "source": [
    "test_matches_s2v = []\n",
    "\n",
    "cosine_thresholds = np.arange(0,1,0.05)\n",
    "\n",
    "for threshold in cosine_thresholds:\n",
    "    print(\"Checking matches for S2V score >\", threshold)\n",
    "    test_matches = []\n",
    "    \n",
    "    for ID in range(len(documents_query_s2v)):\n",
    "\n",
    "        # Scenario 2: mass + sort by Spec2Vec\n",
    "        df_select = found_matches_s2v[ID][(found_matches_s2v[ID]['mass_match'] == 1)\n",
    "                                      & (found_matches_s2v[ID]['s2v_score'] > threshold)]\n",
    "        if df_select.shape[0] > 0:\n",
    "            best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).index[0]\n",
    "            inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "            inchikey_query = documents_query_s2v[ID]._obj.get(\"inchikey\")[:14]\n",
    "            \n",
    "            best_bet = 1 * (inchikey_selected == inchikey_query)\n",
    "        else:\n",
    "            best_bet = -1 # meaning: not found\n",
    "        test_matches.append(best_bet)\n",
    "\n",
    "    # Make arrays from lists:\n",
    "    test_arr = np.array(test_matches)\n",
    "    \n",
    "    test_matches_s2v.append([np.sum(test_arr == 1), np.sum(test_arr == 0), np.sum(test_arr == -1)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'true positives')"
      ]
     },
     "execution_count": 104,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 504x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "test_matches_s2v_arr = np.array(test_matches_s2v)\n",
    "\n",
    "plt.figure(figsize=(7,6))\n",
    "plt.style.use('ggplot')\n",
    "plt.scatter(test_matches_s2v_arr[:,1], test_matches_s2v_arr[:,0])\n",
    "plt.xlabel('false positives')\n",
    "plt.ylabel('true positives')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[790, 209,   1],\n",
       "       [790, 209,   1],\n",
       "       [790, 207,   3],\n",
       "       [790, 200,  10],\n",
       "       [790, 193,  17],\n",
       "       [789, 182,  29],\n",
       "       [784, 167,  49],\n",
       "       [778, 160,  62],\n",
       "       [772, 149,  79],\n",
       "       [764, 143,  93],\n",
       "       [757, 140, 103],\n",
       "       [748, 135, 117],\n",
       "       [740, 128, 132],\n",
       "       [721, 122, 157],\n",
       "       [699, 107, 194],\n",
       "       [657,  98, 245],\n",
       "       [618,  88, 294],\n",
       "       [536,  79, 385],\n",
       "       [419,  71, 510],\n",
       "       [213,  42, 745]])"
      ]
     },
     "execution_count": 105,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.array(test_matches_s2v)[:,:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Checking matches for S2V score > 0.4\n"
     ]
    }
   ],
   "source": [
    "#Trying to find a label of true or false positive for each query at a certain s2v score threshold\n",
    "true_pos_s2v = []\n",
    "false_pos_s2v = []\n",
    "#df ends up as rows: query document number, cols: features (from found_matches) and true or false positive\n",
    "best_matches_labeled_s2v = pd.DataFrame()\n",
    "\n",
    "threshold = 0.4\n",
    "print(\"Checking matches for S2V score >\", threshold)\n",
    "\n",
    "for ID in range(len(documents_query_s2v)):\n",
    "    # Scenario 2: mass + sort by Spec2Vec\n",
    "    df_select = found_matches_s2v[ID][(found_matches_s2v[ID]['mass_match'] == 1)\n",
    "                                  & (found_matches_s2v[ID]['s2v_score'] > threshold)]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0]\n",
    "        best_match_ID = best_match.name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        inchikey_query = documents_query_s2v[ID]._obj.get(\"inchikey\")[:14]\n",
    "\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)\n",
    "        \n",
    "        best_match['label'] = best_bet\n",
    "        best_match.rename(ID, inplace=True)\n",
    "        best_matches_labeled_s2v = best_matches_labeled_s2v.append(best_match)\n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found - not interesting in this case\n",
    "\n",
    "    #record pairs\n",
    "    if best_bet == 1:\n",
    "        true_pos_s2v.append((ID, best_match_ID))\n",
    "    elif best_bet == 0:\n",
    "        false_pos_s2v.append((ID, best_match_ID)) #record best hit for false positive"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "772 149\n",
      "(0, 46003)\n",
      "    cosine_matches  cosine_score  label  mass_match  mod_cosine_matches  \\\n",
      "8              4.0      0.474605    0.0         1.0                 4.0   \n",
      "16             0.0      0.000000    0.0         1.0                 0.0   \n",
      "18             7.0      0.000094    0.0         1.0                10.0   \n",
      "22           134.0      0.717124    0.0         1.0               134.0   \n",
      "46             9.0      0.987546    0.0         1.0                 9.0   \n",
      "48            65.0      0.016529    0.0         1.0                72.0   \n",
      "52            65.0      0.999897    0.0         1.0                65.0   \n",
      "65           141.0      0.663807    0.0         1.0               140.0   \n",
      "66            51.0      0.997571    0.0         1.0                51.0   \n",
      "74             2.0      0.504692    0.0         1.0                 2.0   \n",
      "\n",
      "    mod_cosine_score  s2v_score  \n",
      "8           0.474605   0.480207  \n",
      "16          0.000000   0.949601  \n",
      "18          0.000190   0.478899  \n",
      "22          0.717124   0.950526  \n",
      "46          0.987546   0.980149  \n",
      "48          0.020745   0.439948  \n",
      "52          0.999912   0.997198  \n",
      "65          0.663793   0.747978  \n",
      "66          0.997571   0.983277  \n",
      "74          0.504692   0.676996  \n"
     ]
    }
   ],
   "source": [
    "print(len(true_pos_s2v), len(false_pos_s2v))\n",
    "print(true_pos_s2v[0])\n",
    "print(best_matches_labeled_s2v[best_matches_labeled_s2v['label']==0].iloc[0:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [],
   "source": [
    "#store labels\n",
    "labels = np.array(best_matches_labeled_s2v['label'])\n",
    "#remove labels\n",
    "best_matches_s2v = best_matches_labeled_s2v.drop('label', axis = 1)\n",
    "feature_list = list(best_matches_s2v.columns)\n",
    "best_matches_s2v = np.array(best_matches_s2v)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Features Shape: (690, 6)\n",
      "Training Labels Shape: (690,)\n",
      "Testing Features Shape: (231, 6)\n",
      "Testing Labels Shape: (231,)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "#make training and testing sets\n",
    "train_features, test_features,\\\n",
    "train_labels, test_labels = train_test_split(best_matches_s2v, labels,\n",
    "                                             test_size = 0.25, random_state = 42)\n",
    "print('Training Features Shape:', train_features.shape)\n",
    "print('Training Labels Shape:', train_labels.shape)\n",
    "print('Testing Features Shape:', test_features.shape)\n",
    "print('Testing Labels Shape:', test_labels.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0. 1. 0. 1. 1. 1. 1. 1. 1. 1. 1. 0. 1. 1. 1. 1. 1. 0. 1. 1. 0. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 0. 0. 1. 1. 0. 1. 1. 1. 1. 0. 1. 1. 0. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 0. 1. 1. 0. 1. 0. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 0. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 0. 1. 1. 1. 1. 1. 1. 1. 1. 0. 0. 0. 1. 1. 1.\n",
      " 0. 1. 1. 1. 0. 1. 1. 1. 0. 1. 1. 1. 1. 1. 0. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 0. 1. 0. 1. 1. 1. 0. 1. 1. 1. 1. 1. 1. 1. 0. 1. 0. 1. 1. 1. 1. 1. 0. 0.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 0. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 0. 1. 1. 1. 0. 1. 1. 1. 1. 1. 1. 0. 1. 1. 1. 1. 1. 0. 1. 0. 1. 1. 1. 1.\n",
      " 0. 1. 1. 1. 1. 1. 1. 1. 1. 0. 0. 1. 1. 1. 1.]\n",
      "False positive rate test set:  0.16450216450216448\n"
     ]
    }
   ],
   "source": [
    "print(test_labels)\n",
    "print(\"False positive rate test set: \", 1 - (np.sum(test_labels)/len(test_labels)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(n_estimators=1000, random_state=42)"
      ]
     },
     "execution_count": 111,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "rf = RandomForestClassifier(n_estimators = 1000, random_state = 42)\n",
    "rf.fit(train_features, train_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 0. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 0. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 0. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 0. 1. 1. 1. 1. 1. 1. 1. 1. 0. 1. 1. 1. 0. 0. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.]\n",
      "Accuracy: 0.8311688311688312\n",
      "False positive rate predictions:  0.030303030303030276\n"
     ]
    }
   ],
   "source": [
    "predictions = rf.predict(test_features)\n",
    "print(predictions)\n",
    "accuracy = sum(1 for test, pred in zip(test_labels, predictions) if test == pred) / len(test_labels)\n",
    "print(\"Accuracy:\", accuracy)\n",
    "print(\"False positive rate predictions: \", 1 - (np.sum(predictions)/len(predictions)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Variable: s2v_score            Importance: 0.25\n",
      "Variable: cosine_score         Importance: 0.21\n",
      "Variable: mod_cosine_score     Importance: 0.2\n",
      "Variable: cosine_matches       Importance: 0.17\n",
      "Variable: mod_cosine_matches   Importance: 0.17\n",
      "Variable: mass_match           Importance: 0.0\n"
     ]
    }
   ],
   "source": [
    "# Get numerical feature importances\n",
    "importances = list(rf.feature_importances_)\n",
    "# List of tuples with variable and importance\n",
    "feature_importances = [(feature, round(importance, 2)) for feature, importance in zip(feature_list, importances)]\n",
    "# Sort the feature importances by most important first\n",
    "feature_importances = sorted(feature_importances, key = lambda x: x[1], reverse = True)\n",
    "# Print out the feature and importances \n",
    "[print('Variable: {:20} Importance: {}'.format(*pair)) for pair in feature_importances];"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "def do_random_forest(feature_df, label_name, rf_regr = False, verbose = True, num_est = 1000):\n",
    "    '''Split data in training and test, train rf, predict on test\n",
    "\n",
    "    feature_df: pandas DataFrame\n",
    "    label_name: str, name of the label column in feature_df\n",
    "    rf_refr: bool, make a RandomForestRegressor model instead of Classifier\n",
    "    verbose: bool, print more info\n",
    "    num_est: int, n_estimators in RandomForestClassifiers\n",
    "    \n",
    "    Returns test_features, test_labels, predictions, rf\n",
    "    '''\n",
    "    #store labels\n",
    "    labels = np.array(feature_df[label_name])\n",
    "    #remove labels\n",
    "    feature_df_nolab = feature_df.drop(label_name, axis = 1)\n",
    "    feature_list = list(feature_df_nolab.columns)\n",
    "    feature_df_nolab = np.array(feature_df_nolab)\n",
    "    \n",
    "    #split data in training and test set\n",
    "    train_features, test_features,\\\n",
    "    train_labels, test_labels = train_test_split(feature_df_nolab, labels,\n",
    "                                                 test_size = 0.25, random_state = 42)\n",
    "    if verbose:\n",
    "        print('Label distribution:', dict(zip(*np.unique(np.round(labels,1), return_counts=True))))\n",
    "        print('Training Features Shape:', train_features.shape)\n",
    "        print('Training Labels Shape:', train_labels.shape)\n",
    "        print('Training label distribution:', dict(zip(*np.unique(np.round(train_labels,1), return_counts=True))))\n",
    "        print('Testing Features Shape:', test_features.shape)\n",
    "        print('Testing Labels Shape:', test_labels.shape)\n",
    "        print('Test label distribution:', dict(zip(*np.unique(np.round(test_labels,1), return_counts=True))))\n",
    "    \n",
    "    #train model\n",
    "    if not rf_regr:\n",
    "        rf = RandomForestClassifier(n_estimators = num_est, random_state = 42)\n",
    "    else: #make a rf regression model instead of classifier\n",
    "        rf = RandomForestRegressor(n_estimators = num_est, random_state = 42)\n",
    "    rf.fit(train_features, train_labels)\n",
    "    \n",
    "    #predict test set\n",
    "    predictions = rf.predict(test_features)\n",
    "\n",
    "    print('\\nPredicted label distribution:', dict(zip(*np.unique(np.round(predictions,1), return_counts=True))))\n",
    "    df = pd.DataFrame({'Actual': test_labels, 'Predicted': predictions})\n",
    "    print('Predictions:','\\n0->0','1->1')\n",
    "    zero_to_zero = np.sum((df.Actual < 0.5) & (df.Predicted < 0.5))\n",
    "    one_to_one = np.sum((df.Actual >= 0.5) & (df.Predicted >= 0.5))\n",
    "    print(zero_to_zero, one_to_one)\n",
    "    print('0->1','1->0')\n",
    "    print(np.sum((df.Actual < 0.5) & (df.Predicted >= 0.5)), \n",
    "    np.sum((df.Actual >= 0.5) & (df.Predicted < 0.5)))\n",
    "    print(\"Accuracy: {:.2f}\\n\".format((zero_to_zero + one_to_one)/len(predictions)))\n",
    "\n",
    "    #get feature importances\n",
    "    importances = list(rf.feature_importances_)\n",
    "    #list of tuples with variable and importance\n",
    "    feature_importances = [(feature, round(importance, 2)) for feature, importance in zip(feature_list, importances)]\n",
    "    feature_importances = sorted(feature_importances, key = lambda x: x[1], reverse = True)\n",
    "    [print('Variable: {:20} Importance: {}'.format(*pair)) for pair in feature_importances];\n",
    "    \n",
    "    return (test_features, test_labels, predictions, rf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [],
   "source": [
    "def find_inchikey_matches(matches, documents_library, documents_query):\n",
    "    '''To each df in matches, add label for matching inchikey (1) or non-matching (0)\n",
    "\n",
    "    matches: list of pandas DataFrames, library matching result of query on library\n",
    "    documents_library: list of SpectrumDocuments, spectra in library\n",
    "    documents_query: list of SpectrumDocuments, spectra in query set.\n",
    "        Indices should correspond to indices of matches.\n",
    "    matches_with_labels: list of pandas DataFrames, library matching result of query\n",
    "        on library with matching labels\n",
    "    '''\n",
    "    matches_with_labels = []\n",
    "    for query_id in range(len(documents_query)):\n",
    "        df = matches[query_id].copy()\n",
    "        library_ids = df.index.values\n",
    "        labels = []\n",
    "        query_inchi = documents_query[query_id]._obj.get(\"inchikey\")[:14]\n",
    "        for lib_id in library_ids:\n",
    "            lib_inchi = documents_library[lib_id]._obj.get(\"inchikey\")[:14]\n",
    "            lab = 0\n",
    "            if query_inchi == lib_inchi:\n",
    "                lab = 1\n",
    "            labels.append(lab)\n",
    "        df['label'] = labels\n",
    "        matches_with_labels.append(df)\n",
    "    return matches_with_labels\n",
    "\n",
    "found_matches_s2v_labels = find_inchikey_matches(found_matches_s2v, documents_library_s2v, documents_query_s2v)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Take all hits and label true positives"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>6091</th>\n",
       "      <td>0.991375</td>\n",
       "      <td>4</td>\n",
       "      <td>0.991375</td>\n",
       "      <td>4</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.940499</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6100</th>\n",
       "      <td>0.444585</td>\n",
       "      <td>3</td>\n",
       "      <td>0.444585</td>\n",
       "      <td>3</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.896288</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6092</th>\n",
       "      <td>0.345093</td>\n",
       "      <td>4</td>\n",
       "      <td>0.345093</td>\n",
       "      <td>4</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.874877</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>59577</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.130303</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.741305</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>59660</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.105543</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.738891</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24360</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.188801</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.072580</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24038</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.158695</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.075708</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9790</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.077540</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3256</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.078467</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6013</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.139723</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>329 rows × 7 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "6091       0.991375               4          0.991375                   4   \n",
       "6100       0.444585               3          0.444585                   3   \n",
       "6092       0.345093               4          0.345093                   4   \n",
       "59577      0.000000               0          0.130303                   1   \n",
       "59660      0.000000               0          0.105543                   1   \n",
       "...             ...             ...               ...                 ...   \n",
       "24360      0.000000               0          0.188801                   1   \n",
       "24038      0.000000               0          0.158695                   1   \n",
       "9790       0.000000               0          0.000000                   0   \n",
       "3256       0.000000               0          0.000000                   0   \n",
       "6013       0.000000               0          0.000000                   0   \n",
       "\n",
       "       mass_match  s2v_score  label  \n",
       "6091          1.0   0.940499      1  \n",
       "6100          1.0   0.896288      1  \n",
       "6092          1.0   0.874877      1  \n",
       "59577         0.0   0.741305      0  \n",
       "59660         0.0   0.738891      0  \n",
       "...           ...        ...    ...  \n",
       "24360         1.0  -0.072580      0  \n",
       "24038         1.0  -0.075708      0  \n",
       "9790          1.0  -0.077540      0  \n",
       "3256          1.0  -0.078467      0  \n",
       "6013          1.0  -0.139723      0  \n",
       "\n",
       "[329 rows x 7 columns]"
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "found_matches_s2v_labels[1].sort_values(\"s2v_score\", ascending = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(239233, 7) 239233\n"
     ]
    }
   ],
   "source": [
    "all_matches_s2v_labels = found_matches_s2v_labels[0].append(found_matches_s2v_labels[1:])\n",
    "print(all_matches_s2v_labels.shape, sum(df.shape[0] for df in found_matches_s2v_labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label distribution: {0: 228606, 1: 10627}\n",
      "Training Features Shape: (179424, 6)\n",
      "Training Labels Shape: (179424,)\n",
      "Training label distribution: {0: 171430, 1: 7994}\n",
      "Testing Features Shape: (59809, 6)\n",
      "Testing Labels Shape: (59809,)\n",
      "Test label distribution: {0: 57176, 1: 2633}\n",
      "\n",
      "Predicted label distribution: {0: 57827, 1: 1982}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "56247 1053\n",
      "0->1 1->0\n",
      "929 1580\n",
      "Accuracy: 0.96\n",
      "\n",
      "Variable: s2v_score            Importance: 0.46\n",
      "Variable: cosine_score         Importance: 0.18\n",
      "Variable: mod_cosine_score     Importance: 0.16\n",
      "Variable: mod_cosine_matches   Importance: 0.07\n",
      "Variable: mass_match           Importance: 0.07\n",
      "Variable: cosine_matches       Importance: 0.06\n"
     ]
    }
   ],
   "source": [
    "all_result = do_random_forest(all_matches_s2v_labels, 'label', verbose = True, num_est = 200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {},
   "outputs": [],
   "source": [
    "def make_s2v_selection(matches, topn):\n",
    "    '''Return one df with topn matches (based on s2v_score) for each query\n",
    "    \n",
    "    matches: list of pandas DataFrames, library matching result with matching labels\n",
    "    topn: int, number of top hits to select, sorted on s2v_score\n",
    "    all_selected_matches: pandas DataFrames, concatonated selection of matches for each query\n",
    "    '''\n",
    "    selected_matches = []\n",
    "    for df in matches:\n",
    "        selected_matches.append(df.sort_values(by=['s2v_score'], ascending=False).iloc[:topn])\n",
    "    all_selected_matches = selected_matches[0].append(selected_matches[1:])\n",
    "    return all_selected_matches\n",
    "\n",
    "top20_found_matches_s2v_labels = make_s2v_selection(found_matches_s2v_labels, 20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>46003</th>\n",
       "      <td>0.960780</td>\n",
       "      <td>3</td>\n",
       "      <td>0.960780</td>\n",
       "      <td>3</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.887800</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38195</th>\n",
       "      <td>0.049197</td>\n",
       "      <td>1</td>\n",
       "      <td>0.971344</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.858902</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48062</th>\n",
       "      <td>0.043040</td>\n",
       "      <td>2</td>\n",
       "      <td>0.043040</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.848561</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45969</th>\n",
       "      <td>0.049498</td>\n",
       "      <td>1</td>\n",
       "      <td>0.961919</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.844523</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32755</th>\n",
       "      <td>0.119678</td>\n",
       "      <td>1</td>\n",
       "      <td>0.608107</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.811812</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41449</th>\n",
       "      <td>0.999194</td>\n",
       "      <td>40</td>\n",
       "      <td>0.999194</td>\n",
       "      <td>40</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.979332</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45245</th>\n",
       "      <td>0.999585</td>\n",
       "      <td>38</td>\n",
       "      <td>0.999585</td>\n",
       "      <td>38</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.979218</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44988</th>\n",
       "      <td>0.999373</td>\n",
       "      <td>35</td>\n",
       "      <td>0.999373</td>\n",
       "      <td>35</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.979008</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54877</th>\n",
       "      <td>0.999565</td>\n",
       "      <td>34</td>\n",
       "      <td>0.999565</td>\n",
       "      <td>34</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.978979</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>51669</th>\n",
       "      <td>0.999435</td>\n",
       "      <td>38</td>\n",
       "      <td>0.999435</td>\n",
       "      <td>38</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.978883</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>20000 rows × 7 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "46003      0.960780               3          0.960780                   3   \n",
       "38195      0.049197               1          0.971344                   3   \n",
       "48062      0.043040               2          0.043040                   2   \n",
       "45969      0.049498               1          0.961919                   2   \n",
       "32755      0.119678               1          0.608107                   2   \n",
       "...             ...             ...               ...                 ...   \n",
       "41449      0.999194              40          0.999194                  40   \n",
       "45245      0.999585              38          0.999585                  38   \n",
       "44988      0.999373              35          0.999373                  35   \n",
       "54877      0.999565              34          0.999565                  34   \n",
       "51669      0.999435              38          0.999435                  38   \n",
       "\n",
       "       mass_match  s2v_score  label  \n",
       "46003         1.0   0.887800      1  \n",
       "38195         0.0   0.858902      0  \n",
       "48062         0.0   0.848561      0  \n",
       "45969         0.0   0.844523      0  \n",
       "32755         0.0   0.811812      0  \n",
       "...           ...        ...    ...  \n",
       "41449         0.0   0.979332      0  \n",
       "45245         0.0   0.979218      0  \n",
       "44988         0.0   0.979008      0  \n",
       "54877         0.0   0.978979      0  \n",
       "51669         0.0   0.978883      0  \n",
       "\n",
       "[20000 rows x 7 columns]"
      ]
     },
     "execution_count": 120,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "top20_found_matches_s2v_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label distribution: {0: 16122, 1: 3878}\n",
      "Training Features Shape: (15000, 6)\n",
      "Training Labels Shape: (15000,)\n",
      "Training label distribution: {0: 12115, 1: 2885}\n",
      "Testing Features Shape: (5000, 6)\n",
      "Testing Labels Shape: (5000,)\n",
      "Test label distribution: {0: 4007, 1: 993}\n",
      "\n",
      "Predicted label distribution: {0: 4190, 1: 810}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "3805 608\n",
      "0->1 1->0\n",
      "202 385\n",
      "Accuracy: 0.88\n",
      "\n",
      "Variable: mass_match           Importance: 0.28\n",
      "Variable: s2v_score            Importance: 0.21\n",
      "Variable: cosine_score         Importance: 0.15\n",
      "Variable: mod_cosine_score     Importance: 0.14\n",
      "Variable: cosine_matches       Importance: 0.11\n",
      "Variable: mod_cosine_matches   Importance: 0.11\n"
     ]
    }
   ],
   "source": [
    "results_top20 = do_random_forest(top20_found_matches_s2v_labels, 'label', verbose = True, num_est = 500)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>label</th>\n",
       "      <th>prediction</th>\n",
       "      <th>correct_guess</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.703666</td>\n",
       "      <td>16.0</td>\n",
       "      <td>0.703666</td>\n",
       "      <td>16.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.861758</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.911023</td>\n",
       "      <td>102.0</td>\n",
       "      <td>0.911231</td>\n",
       "      <td>102.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.809915</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.644169</td>\n",
       "      <td>110.0</td>\n",
       "      <td>0.644786</td>\n",
       "      <td>111.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.723040</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.023764</td>\n",
       "      <td>12.0</td>\n",
       "      <td>0.023767</td>\n",
       "      <td>13.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.467888</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.731167</td>\n",
       "      <td>82.0</td>\n",
       "      <td>0.731167</td>\n",
       "      <td>82.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.677396</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4962</th>\n",
       "      <td>0.122735</td>\n",
       "      <td>172.0</td>\n",
       "      <td>0.122735</td>\n",
       "      <td>172.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.688703</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4966</th>\n",
       "      <td>0.637036</td>\n",
       "      <td>49.0</td>\n",
       "      <td>0.637036</td>\n",
       "      <td>49.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.663187</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4967</th>\n",
       "      <td>0.631012</td>\n",
       "      <td>33.0</td>\n",
       "      <td>0.631019</td>\n",
       "      <td>34.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.731506</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4968</th>\n",
       "      <td>0.329908</td>\n",
       "      <td>74.0</td>\n",
       "      <td>0.370854</td>\n",
       "      <td>83.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.631261</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4969</th>\n",
       "      <td>0.900187</td>\n",
       "      <td>56.0</td>\n",
       "      <td>0.900187</td>\n",
       "      <td>56.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.852985</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>587 rows × 9 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "0         0.703666            16.0          0.703666                16.0   \n",
       "1         0.911023           102.0          0.911231               102.0   \n",
       "4         0.644169           110.0          0.644786               111.0   \n",
       "11        0.023764            12.0          0.023767                13.0   \n",
       "12        0.731167            82.0          0.731167                82.0   \n",
       "...            ...             ...               ...                 ...   \n",
       "4962      0.122735           172.0          0.122735               172.0   \n",
       "4966      0.637036            49.0          0.637036                49.0   \n",
       "4967      0.631012            33.0          0.631019                34.0   \n",
       "4968      0.329908            74.0          0.370854                83.0   \n",
       "4969      0.900187            56.0          0.900187                56.0   \n",
       "\n",
       "      mass_match  s2v_score  label  prediction  correct_guess  \n",
       "0            1.0   0.861758      0           1          False  \n",
       "1            0.0   0.809915      1           0          False  \n",
       "4            0.0   0.723040      1           0          False  \n",
       "11           0.0   0.467888      1           0          False  \n",
       "12           1.0   0.677396      1           0          False  \n",
       "...          ...        ...    ...         ...            ...  \n",
       "4962         0.0   0.688703      1           0          False  \n",
       "4966         0.0   0.663187      1           0          False  \n",
       "4967         1.0   0.731506      0           1          False  \n",
       "4968         0.0   0.631261      1           0          False  \n",
       "4969         1.0   0.852985      1           0          False  \n",
       "\n",
       "[587 rows x 9 columns]"
      ]
     },
     "execution_count": 122,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cols = ['cosine_score','cosine_matches','mod_cosine_score','mod_cosine_matches','mass_match','s2v_score']\n",
    "df_results_top20 = pd.DataFrame(results_top20[0], columns = cols)\n",
    "df_results_top20['label'] = results_top20[1]\n",
    "df_results_top20['prediction'] = results_top20[2]\n",
    "df_results_top20['correct_guess'] = (df_results_top20['label'] == df_results_top20['prediction'])\n",
    "df_results_top20[df_results_top20['correct_guess'] == False]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "215\n",
      "385\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>label</th>\n",
       "      <th>prediction</th>\n",
       "      <th>correct_guess</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.911023</td>\n",
       "      <td>102.0</td>\n",
       "      <td>0.911231</td>\n",
       "      <td>102.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.809915</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.644169</td>\n",
       "      <td>110.0</td>\n",
       "      <td>0.644786</td>\n",
       "      <td>111.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.723040</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.023764</td>\n",
       "      <td>12.0</td>\n",
       "      <td>0.023767</td>\n",
       "      <td>13.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.467888</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.731167</td>\n",
       "      <td>82.0</td>\n",
       "      <td>0.731167</td>\n",
       "      <td>82.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.677396</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>0.481067</td>\n",
       "      <td>6.0</td>\n",
       "      <td>0.481067</td>\n",
       "      <td>6.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.532788</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>0.585359</td>\n",
       "      <td>38.0</td>\n",
       "      <td>0.585359</td>\n",
       "      <td>38.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.668694</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>51</th>\n",
       "      <td>0.333450</td>\n",
       "      <td>71.0</td>\n",
       "      <td>0.338706</td>\n",
       "      <td>77.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.626964</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>60</th>\n",
       "      <td>0.064025</td>\n",
       "      <td>34.0</td>\n",
       "      <td>0.065043</td>\n",
       "      <td>36.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.457384</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>81</th>\n",
       "      <td>0.975349</td>\n",
       "      <td>18.0</td>\n",
       "      <td>0.975349</td>\n",
       "      <td>18.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.704903</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>116</th>\n",
       "      <td>0.662041</td>\n",
       "      <td>13.0</td>\n",
       "      <td>0.682216</td>\n",
       "      <td>16.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.583011</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>120</th>\n",
       "      <td>0.993707</td>\n",
       "      <td>32.0</td>\n",
       "      <td>0.993727</td>\n",
       "      <td>33.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.538273</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>137</th>\n",
       "      <td>0.803508</td>\n",
       "      <td>24.0</td>\n",
       "      <td>0.805839</td>\n",
       "      <td>24.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.580528</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>202</th>\n",
       "      <td>0.682805</td>\n",
       "      <td>134.0</td>\n",
       "      <td>0.683565</td>\n",
       "      <td>135.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.762182</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>231</th>\n",
       "      <td>0.496451</td>\n",
       "      <td>56.0</td>\n",
       "      <td>0.496451</td>\n",
       "      <td>56.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.644734</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>242</th>\n",
       "      <td>0.235687</td>\n",
       "      <td>8.0</td>\n",
       "      <td>0.235687</td>\n",
       "      <td>8.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.755786</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>252</th>\n",
       "      <td>0.364518</td>\n",
       "      <td>81.0</td>\n",
       "      <td>0.364518</td>\n",
       "      <td>81.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.637288</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>259</th>\n",
       "      <td>0.962775</td>\n",
       "      <td>89.0</td>\n",
       "      <td>0.962775</td>\n",
       "      <td>89.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.832009</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>272</th>\n",
       "      <td>0.712328</td>\n",
       "      <td>153.0</td>\n",
       "      <td>0.712259</td>\n",
       "      <td>158.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.776626</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>290</th>\n",
       "      <td>0.794071</td>\n",
       "      <td>34.0</td>\n",
       "      <td>0.794071</td>\n",
       "      <td>34.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.727486</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>353</th>\n",
       "      <td>0.000585</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.252320</td>\n",
       "      <td>8.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.487066</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "1        0.911023           102.0          0.911231               102.0   \n",
       "4        0.644169           110.0          0.644786               111.0   \n",
       "11       0.023764            12.0          0.023767                13.0   \n",
       "12       0.731167            82.0          0.731167                82.0   \n",
       "27       0.481067             6.0          0.481067                 6.0   \n",
       "34       0.585359            38.0          0.585359                38.0   \n",
       "51       0.333450            71.0          0.338706                77.0   \n",
       "60       0.064025            34.0          0.065043                36.0   \n",
       "81       0.975349            18.0          0.975349                18.0   \n",
       "116      0.662041            13.0          0.682216                16.0   \n",
       "120      0.993707            32.0          0.993727                33.0   \n",
       "137      0.803508            24.0          0.805839                24.0   \n",
       "202      0.682805           134.0          0.683565               135.0   \n",
       "231      0.496451            56.0          0.496451                56.0   \n",
       "242      0.235687             8.0          0.235687                 8.0   \n",
       "252      0.364518            81.0          0.364518                81.0   \n",
       "259      0.962775            89.0          0.962775                89.0   \n",
       "272      0.712328           153.0          0.712259               158.0   \n",
       "290      0.794071            34.0          0.794071                34.0   \n",
       "353      0.000585             2.0          0.252320                 8.0   \n",
       "\n",
       "     mass_match  s2v_score  label  prediction  correct_guess  \n",
       "1           0.0   0.809915      1           0          False  \n",
       "4           0.0   0.723040      1           0          False  \n",
       "11          0.0   0.467888      1           0          False  \n",
       "12          1.0   0.677396      1           0          False  \n",
       "27          1.0   0.532788      1           0          False  \n",
       "34          0.0   0.668694      1           0          False  \n",
       "51          0.0   0.626964      1           0          False  \n",
       "60          0.0   0.457384      1           0          False  \n",
       "81          0.0   0.704903      1           0          False  \n",
       "116         1.0   0.583011      1           0          False  \n",
       "120         0.0   0.538273      1           0          False  \n",
       "137         0.0   0.580528      1           0          False  \n",
       "202         0.0   0.762182      1           0          False  \n",
       "231         1.0   0.644734      1           0          False  \n",
       "242         0.0   0.755786      1           0          False  \n",
       "252         1.0   0.637288      1           0          False  \n",
       "259         0.0   0.832009      1           0          False  \n",
       "272         0.0   0.776626      1           0          False  \n",
       "290         1.0   0.727486      1           0          False  \n",
       "353         0.0   0.487066      1           0          False  "
      ]
     },
     "execution_count": 123,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#how many 1->0 predictions can be explained by absent mass_match? --> 214/386\n",
    "print(df_results_top20[(df_results_top20['label'] == 1) & (df_results_top20['prediction'] == 0) & (df_results_top20['mass_match'] == 0)].shape[0])\n",
    "print(df_results_top20[(df_results_top20['label'] == 1) & (df_results_top20['prediction'] == 0)].shape[0])\n",
    "df_results_top20[(df_results_top20['label'] == 1) & (df_results_top20['prediction'] == 0)].iloc[:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "182\n",
      "202\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>label</th>\n",
       "      <th>prediction</th>\n",
       "      <th>correct_guess</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.703666</td>\n",
       "      <td>16.0</td>\n",
       "      <td>0.703666</td>\n",
       "      <td>16.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.861758</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>0.830622</td>\n",
       "      <td>193.0</td>\n",
       "      <td>0.830622</td>\n",
       "      <td>193.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.906160</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>0.349962</td>\n",
       "      <td>7.0</td>\n",
       "      <td>0.349962</td>\n",
       "      <td>7.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.486539</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>70</th>\n",
       "      <td>0.483246</td>\n",
       "      <td>74.0</td>\n",
       "      <td>0.483259</td>\n",
       "      <td>75.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.804983</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89</th>\n",
       "      <td>0.949021</td>\n",
       "      <td>16.0</td>\n",
       "      <td>0.949249</td>\n",
       "      <td>20.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.884175</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>92</th>\n",
       "      <td>0.759489</td>\n",
       "      <td>158.0</td>\n",
       "      <td>0.759489</td>\n",
       "      <td>158.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.877990</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>114</th>\n",
       "      <td>0.606647</td>\n",
       "      <td>13.0</td>\n",
       "      <td>0.606647</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.937617</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>129</th>\n",
       "      <td>0.135855</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.135855</td>\n",
       "      <td>30.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.526047</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>161</th>\n",
       "      <td>0.924346</td>\n",
       "      <td>3.0</td>\n",
       "      <td>0.924346</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.712865</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178</th>\n",
       "      <td>0.980240</td>\n",
       "      <td>7.0</td>\n",
       "      <td>0.980240</td>\n",
       "      <td>7.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.913873</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>244</th>\n",
       "      <td>0.804156</td>\n",
       "      <td>18.0</td>\n",
       "      <td>0.804167</td>\n",
       "      <td>19.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.899833</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>257</th>\n",
       "      <td>0.847378</td>\n",
       "      <td>19.0</td>\n",
       "      <td>0.847378</td>\n",
       "      <td>19.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.705288</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>271</th>\n",
       "      <td>0.922820</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.922820</td>\n",
       "      <td>15.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.784235</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>305</th>\n",
       "      <td>0.463891</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.463891</td>\n",
       "      <td>15.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.877230</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>306</th>\n",
       "      <td>0.917684</td>\n",
       "      <td>94.0</td>\n",
       "      <td>0.917684</td>\n",
       "      <td>94.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.940612</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>339</th>\n",
       "      <td>0.597664</td>\n",
       "      <td>10.0</td>\n",
       "      <td>0.597664</td>\n",
       "      <td>10.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.617402</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>341</th>\n",
       "      <td>0.709450</td>\n",
       "      <td>72.0</td>\n",
       "      <td>0.709450</td>\n",
       "      <td>72.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.843570</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>372</th>\n",
       "      <td>0.897057</td>\n",
       "      <td>18.0</td>\n",
       "      <td>0.897057</td>\n",
       "      <td>18.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.818866</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>381</th>\n",
       "      <td>0.460082</td>\n",
       "      <td>87.0</td>\n",
       "      <td>0.467795</td>\n",
       "      <td>88.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.734475</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>408</th>\n",
       "      <td>0.622711</td>\n",
       "      <td>82.0</td>\n",
       "      <td>0.622802</td>\n",
       "      <td>83.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.883726</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "0        0.703666            16.0          0.703666                16.0   \n",
       "31       0.830622           193.0          0.830622               193.0   \n",
       "39       0.349962             7.0          0.349962                 7.0   \n",
       "70       0.483246            74.0          0.483259                75.0   \n",
       "89       0.949021            16.0          0.949249                20.0   \n",
       "92       0.759489           158.0          0.759489               158.0   \n",
       "114      0.606647            13.0          0.606647                13.0   \n",
       "129      0.135855            30.0          0.135855                30.0   \n",
       "161      0.924346             3.0          0.924346                 3.0   \n",
       "178      0.980240             7.0          0.980240                 7.0   \n",
       "244      0.804156            18.0          0.804167                19.0   \n",
       "257      0.847378            19.0          0.847378                19.0   \n",
       "271      0.922820            15.0          0.922820                15.0   \n",
       "305      0.463891            15.0          0.463891                15.0   \n",
       "306      0.917684            94.0          0.917684                94.0   \n",
       "339      0.597664            10.0          0.597664                10.0   \n",
       "341      0.709450            72.0          0.709450                72.0   \n",
       "372      0.897057            18.0          0.897057                18.0   \n",
       "381      0.460082            87.0          0.467795                88.0   \n",
       "408      0.622711            82.0          0.622802                83.0   \n",
       "\n",
       "     mass_match  s2v_score  label  prediction  correct_guess  \n",
       "0           1.0   0.861758      0           1          False  \n",
       "31          1.0   0.906160      0           1          False  \n",
       "39          1.0   0.486539      0           1          False  \n",
       "70          1.0   0.804983      0           1          False  \n",
       "89          1.0   0.884175      0           1          False  \n",
       "92          1.0   0.877990      0           1          False  \n",
       "114         1.0   0.937617      0           1          False  \n",
       "129         1.0   0.526047      0           1          False  \n",
       "161         1.0   0.712865      0           1          False  \n",
       "178         1.0   0.913873      0           1          False  \n",
       "244         1.0   0.899833      0           1          False  \n",
       "257         1.0   0.705288      0           1          False  \n",
       "271         1.0   0.784235      0           1          False  \n",
       "305         1.0   0.877230      0           1          False  \n",
       "306         0.0   0.940612      0           1          False  \n",
       "339         1.0   0.617402      0           1          False  \n",
       "341         1.0   0.843570      0           1          False  \n",
       "372         1.0   0.818866      0           1          False  \n",
       "381         1.0   0.734475      0           1          False  \n",
       "408         1.0   0.883726      0           1          False  "
      ]
     },
     "execution_count": 124,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#how many 0->1 predictions can be explained by present mass_match? --> 180/202\n",
    "print(df_results_top20[(df_results_top20['label'] == 0) & (df_results_top20['prediction'] == 1) & (df_results_top20['mass_match'] == 1)].shape[0])\n",
    "print(df_results_top20[(df_results_top20['label'] == 0) & (df_results_top20['prediction'] == 1)].shape[0])\n",
    "df_results_top20[(df_results_top20['label'] == 0) & (df_results_top20['prediction'] == 1)].iloc[:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(7756, 7)"
      ]
     },
     "execution_count": 125,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#make ratio false to true positives 50/50\n",
    "true_p = top20_found_matches_s2v_labels[top20_found_matches_s2v_labels['label']==1]\n",
    "num_true_p = true_p.shape[0]\n",
    "np.random.seed(42) # to make it reproducible\n",
    "chosen_false_p_names = np.random.choice(range(top20_found_matches_s2v_labels[top20_found_matches_s2v_labels['label']==0].shape[0]), num_true_p)\n",
    "top20_found_matches_s2v_labels_5050 = true_p.append(top20_found_matches_s2v_labels[top20_found_matches_s2v_labels['label']==0].iloc[chosen_false_p_names])\n",
    "top20_found_matches_s2v_labels_5050.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "New shape: (457212, 7)\n"
     ]
    }
   ],
   "source": [
    "#oversample true positives to get equal ratio\n",
    "def oversample(matches, label_column):\n",
    "    '''Oversamples the 1 label so that 1 and 0 label are equal in the df\n",
    "\n",
    "    matches: pandas DataFrame, concatenation of all queries matching to a library\n",
    "    label_column: str, the column that contains the label\n",
    "    '''\n",
    "    all_true_p = matches.copy()[matches[label_column]==1]\n",
    "    len_all_true_p = all_true_p.shape[0]\n",
    "    all_false_p = matches.copy()[matches[label_column]==0]\n",
    "    len_all_false_p = all_false_p.shape[0]\n",
    "    ratio = len_all_false_p/len_all_true_p\n",
    "    decimals = ratio % 1\n",
    "    np.random.seed(42)\n",
    "    chosen_indices = np.random.choice(range(len_all_true_p), round(decimals*len_all_true_p))\n",
    "    balanced_true_p = all_true_p.iloc[chosen_indices]\n",
    "    for i in range(1,int(ratio)+1):\n",
    "        balanced_true_p = balanced_true_p.append(all_true_p)\n",
    "    matches_oversampled = balanced_true_p.append(all_false_p)\n",
    "    np.random.seed(42)\n",
    "    matches_oversampled = matches_oversampled.sample(frac=1)\n",
    "    print('New shape:', matches_oversampled.shape)\n",
    "    return matches_oversampled\n",
    "\n",
    "all_matches_s2v_oversampled = oversample(all_matches_s2v_labels, 'label')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label distribution: {0: 228606, 1: 228606}\n",
      "Training Features Shape: (342909, 6)\n",
      "Training Labels Shape: (342909,)\n",
      "Training label distribution: {0: 171294, 1: 171615}\n",
      "Testing Features Shape: (114303, 6)\n",
      "Testing Labels Shape: (114303,)\n",
      "Test label distribution: {0: 57312, 1: 56991}\n",
      "\n",
      "Predicted label distribution: {0: 55912, 1: 58391}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "55820 56899\n",
      "0->1 1->0\n",
      "1492 92\n",
      "Accuracy: 0.99\n",
      "\n",
      "Variable: s2v_score            Importance: 0.48\n",
      "Variable: cosine_score         Importance: 0.19\n",
      "Variable: mod_cosine_score     Importance: 0.15\n",
      "Variable: mod_cosine_matches   Importance: 0.09\n",
      "Variable: cosine_matches       Importance: 0.07\n",
      "Variable: mass_match           Importance: 0.03\n"
     ]
    }
   ],
   "source": [
    "results_all_oversampled = do_random_forest(all_matches_s2v_oversampled, 'label', verbose = True, num_est = 200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label distribution: {0: 3878, 1: 3878}\n",
      "Training Features Shape: (5817, 6)\n",
      "Training Labels Shape: (5817,)\n",
      "Training label distribution: {0: 2911, 1: 2906}\n",
      "Testing Features Shape: (1939, 6)\n",
      "Testing Labels Shape: (1939,)\n",
      "Test label distribution: {0: 967, 1: 972}\n",
      "\n",
      "Predicted label distribution: {0: 1038, 1: 901}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "836 770\n",
      "0->1 1->0\n",
      "131 202\n",
      "Accuracy: 0.83\n",
      "\n",
      "Variable: mass_match           Importance: 0.33\n",
      "Variable: s2v_score            Importance: 0.19\n",
      "Variable: cosine_score         Importance: 0.14\n",
      "Variable: cosine_matches       Importance: 0.12\n",
      "Variable: mod_cosine_score     Importance: 0.11\n",
      "Variable: mod_cosine_matches   Importance: 0.1\n"
     ]
    }
   ],
   "source": [
    "results_top20_5050 = do_random_forest(top20_found_matches_s2v_labels_5050, 'label', verbose = True, num_est = 500)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "166\n",
      "202\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>label</th>\n",
       "      <th>prediction</th>\n",
       "      <th>correct_guess</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.711667</td>\n",
       "      <td>198.0</td>\n",
       "      <td>0.711669</td>\n",
       "      <td>199.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.821747</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.484756</td>\n",
       "      <td>59.0</td>\n",
       "      <td>0.484756</td>\n",
       "      <td>59.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.558689</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.768771</td>\n",
       "      <td>56.0</td>\n",
       "      <td>0.768771</td>\n",
       "      <td>56.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.705430</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>0.417969</td>\n",
       "      <td>28.0</td>\n",
       "      <td>0.418181</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.535101</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>0.365289</td>\n",
       "      <td>6.0</td>\n",
       "      <td>0.366496</td>\n",
       "      <td>8.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.560432</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1905</th>\n",
       "      <td>0.459026</td>\n",
       "      <td>107.0</td>\n",
       "      <td>0.460971</td>\n",
       "      <td>111.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.685349</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1911</th>\n",
       "      <td>0.178072</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.178072</td>\n",
       "      <td>4.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.595440</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1914</th>\n",
       "      <td>0.537589</td>\n",
       "      <td>103.0</td>\n",
       "      <td>0.537589</td>\n",
       "      <td>103.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.598474</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1924</th>\n",
       "      <td>0.152687</td>\n",
       "      <td>27.0</td>\n",
       "      <td>0.154091</td>\n",
       "      <td>34.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.533858</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1933</th>\n",
       "      <td>0.542146</td>\n",
       "      <td>65.0</td>\n",
       "      <td>0.542146</td>\n",
       "      <td>65.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.713790</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>202 rows × 9 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "1         0.711667           198.0          0.711669               199.0   \n",
       "6         0.484756            59.0          0.484756                59.0   \n",
       "11        0.768771            56.0          0.768771                56.0   \n",
       "19        0.417969            28.0          0.418181                30.0   \n",
       "28        0.365289             6.0          0.366496                 8.0   \n",
       "...            ...             ...               ...                 ...   \n",
       "1905      0.459026           107.0          0.460971               111.0   \n",
       "1911      0.178072             4.0          0.178072                 4.0   \n",
       "1914      0.537589           103.0          0.537589               103.0   \n",
       "1924      0.152687            27.0          0.154091                34.0   \n",
       "1933      0.542146            65.0          0.542146                65.0   \n",
       "\n",
       "      mass_match  s2v_score  label  prediction  correct_guess  \n",
       "1            0.0   0.821747      1           0          False  \n",
       "6            0.0   0.558689      1           0          False  \n",
       "11           1.0   0.705430      1           0          False  \n",
       "19           0.0   0.535101      1           0          False  \n",
       "28           0.0   0.560432      1           0          False  \n",
       "...          ...        ...    ...         ...            ...  \n",
       "1905         0.0   0.685349      1           0          False  \n",
       "1911         1.0   0.595440      1           0          False  \n",
       "1914         0.0   0.598474      1           0          False  \n",
       "1924         0.0   0.533858      1           0          False  \n",
       "1933         1.0   0.713790      1           0          False  \n",
       "\n",
       "[202 rows x 9 columns]"
      ]
     },
     "execution_count": 129,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#how many 1->0 predictions can be explained by absent mass_match? --> 162/192\n",
    "df_results_top20_5050 = pd.DataFrame(results_top20_5050[0], columns = cols)\n",
    "df_results_top20_5050['label'] = results_top20_5050[1]\n",
    "df_results_top20_5050['prediction'] = results_top20_5050[2]\n",
    "df_results_top20_5050['correct_guess'] = (df_results_top20_5050['label'] == df_results_top20_5050['prediction'])\n",
    "df_results_top20_5050[df_results_top20_5050['correct_guess'] == False]\n",
    "print(df_results_top20_5050[(df_results_top20_5050['label'] == 1) & (df_results_top20_5050['prediction'] == 0) & (df_results_top20_5050['mass_match'] == 0)].shape[0])\n",
    "print(df_results_top20_5050[(df_results_top20_5050['label'] == 1) & (df_results_top20_5050['prediction'] == 0)].shape[0])\n",
    "df_results_top20_5050[(df_results_top20_5050['label'] == 1) & (df_results_top20_5050['prediction'] == 0)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "87\n",
      "131\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>label</th>\n",
       "      <th>prediction</th>\n",
       "      <th>correct_guess</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0.038221</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.038581</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.555749</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>0.928648</td>\n",
       "      <td>21.0</td>\n",
       "      <td>0.951665</td>\n",
       "      <td>24.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.958238</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52</th>\n",
       "      <td>0.695011</td>\n",
       "      <td>149.0</td>\n",
       "      <td>0.695037</td>\n",
       "      <td>150.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.777567</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>63</th>\n",
       "      <td>0.674875</td>\n",
       "      <td>178.0</td>\n",
       "      <td>0.674884</td>\n",
       "      <td>180.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.823565</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>68</th>\n",
       "      <td>0.445761</td>\n",
       "      <td>118.0</td>\n",
       "      <td>0.445761</td>\n",
       "      <td>118.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.684618</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>82</th>\n",
       "      <td>0.879572</td>\n",
       "      <td>53.0</td>\n",
       "      <td>0.879572</td>\n",
       "      <td>53.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.943423</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99</th>\n",
       "      <td>0.351644</td>\n",
       "      <td>3.0</td>\n",
       "      <td>0.351644</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.790554</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>152</th>\n",
       "      <td>0.687446</td>\n",
       "      <td>46.0</td>\n",
       "      <td>0.687446</td>\n",
       "      <td>46.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.577571</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>180</th>\n",
       "      <td>0.168524</td>\n",
       "      <td>24.0</td>\n",
       "      <td>0.168524</td>\n",
       "      <td>24.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.777927</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>215</th>\n",
       "      <td>0.592601</td>\n",
       "      <td>98.0</td>\n",
       "      <td>0.592601</td>\n",
       "      <td>98.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.774438</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>225</th>\n",
       "      <td>0.611380</td>\n",
       "      <td>99.0</td>\n",
       "      <td>0.615327</td>\n",
       "      <td>101.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.784685</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>233</th>\n",
       "      <td>0.998281</td>\n",
       "      <td>10.0</td>\n",
       "      <td>0.998281</td>\n",
       "      <td>10.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.993132</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>235</th>\n",
       "      <td>0.371712</td>\n",
       "      <td>29.0</td>\n",
       "      <td>0.371779</td>\n",
       "      <td>33.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.657796</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>241</th>\n",
       "      <td>0.526628</td>\n",
       "      <td>73.0</td>\n",
       "      <td>0.526628</td>\n",
       "      <td>73.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.867542</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>265</th>\n",
       "      <td>0.931257</td>\n",
       "      <td>52.0</td>\n",
       "      <td>0.931266</td>\n",
       "      <td>54.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.778754</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>281</th>\n",
       "      <td>0.862007</td>\n",
       "      <td>11.0</td>\n",
       "      <td>0.862007</td>\n",
       "      <td>11.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.654608</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>340</th>\n",
       "      <td>0.976874</td>\n",
       "      <td>12.0</td>\n",
       "      <td>0.976874</td>\n",
       "      <td>12.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.960072</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>373</th>\n",
       "      <td>0.392583</td>\n",
       "      <td>3.0</td>\n",
       "      <td>0.392583</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.520458</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>406</th>\n",
       "      <td>0.597322</td>\n",
       "      <td>91.0</td>\n",
       "      <td>0.597427</td>\n",
       "      <td>92.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.712588</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>417</th>\n",
       "      <td>0.859380</td>\n",
       "      <td>62.0</td>\n",
       "      <td>0.859380</td>\n",
       "      <td>62.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.850937</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "13       0.038221             2.0          0.038581                 3.0   \n",
       "30       0.928648            21.0          0.951665                24.0   \n",
       "52       0.695011           149.0          0.695037               150.0   \n",
       "63       0.674875           178.0          0.674884               180.0   \n",
       "68       0.445761           118.0          0.445761               118.0   \n",
       "82       0.879572            53.0          0.879572                53.0   \n",
       "99       0.351644             3.0          0.351644                 3.0   \n",
       "152      0.687446            46.0          0.687446                46.0   \n",
       "180      0.168524            24.0          0.168524                24.0   \n",
       "215      0.592601            98.0          0.592601                98.0   \n",
       "225      0.611380            99.0          0.615327               101.0   \n",
       "233      0.998281            10.0          0.998281                10.0   \n",
       "235      0.371712            29.0          0.371779                33.0   \n",
       "241      0.526628            73.0          0.526628                73.0   \n",
       "265      0.931257            52.0          0.931266                54.0   \n",
       "281      0.862007            11.0          0.862007                11.0   \n",
       "340      0.976874            12.0          0.976874                12.0   \n",
       "373      0.392583             3.0          0.392583                 3.0   \n",
       "406      0.597322            91.0          0.597427                92.0   \n",
       "417      0.859380            62.0          0.859380                62.0   \n",
       "\n",
       "     mass_match  s2v_score  label  prediction  correct_guess  \n",
       "13          1.0   0.555749      0           1          False  \n",
       "30          0.0   0.958238      0           1          False  \n",
       "52          0.0   0.777567      0           1          False  \n",
       "63          1.0   0.823565      0           1          False  \n",
       "68          1.0   0.684618      0           1          False  \n",
       "82          1.0   0.943423      0           1          False  \n",
       "99          1.0   0.790554      0           1          False  \n",
       "152         0.0   0.577571      0           1          False  \n",
       "180         1.0   0.777927      0           1          False  \n",
       "215         1.0   0.774438      0           1          False  \n",
       "225         0.0   0.784685      0           1          False  \n",
       "233         1.0   0.993132      0           1          False  \n",
       "235         1.0   0.657796      0           1          False  \n",
       "241         1.0   0.867542      0           1          False  \n",
       "265         1.0   0.778754      0           1          False  \n",
       "281         1.0   0.654608      0           1          False  \n",
       "340         1.0   0.960072      0           1          False  \n",
       "373         1.0   0.520458      0           1          False  \n",
       "406         1.0   0.712588      0           1          False  \n",
       "417         0.0   0.850937      0           1          False  "
      ]
     },
     "execution_count": 130,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#how many 0->1 predictions can be explained by present mass_match? --> 88/135\n",
    "print(df_results_top20_5050[(df_results_top20_5050['label'] == 0) & (df_results_top20_5050['prediction'] == 1) & (df_results_top20_5050['mass_match'] == 1)].shape[0])\n",
    "print(df_results_top20_5050[(df_results_top20_5050['label'] == 0) & (df_results_top20_5050['prediction'] == 1)].shape[0])\n",
    "df_results_top20_5050[(df_results_top20_5050['label'] == 0) & (df_results_top20_5050['prediction'] == 1)].iloc[:20]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Plot results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_m_s2v = []\n",
    "test_m_rf_all = []\n",
    "test_m_rf_top20 = []\n",
    "test_m_rf_top20_5050 = []\n",
    "\n",
    "threshold = 0.4\n",
    "\n",
    "\n",
    "for ID in range(len(documents_query_s2v)):\n",
    "    inchikey_query = documents_query_s2v[ID]._obj.get(\"inchikey\")[:14]\n",
    "    current_match = found_matches_s2v_labels[ID].drop('label', axis = 1)\n",
    "\n",
    "    # Scenario 2: mass + sort by Spec2Vec\n",
    "    df_select = current_match[(current_match['mass_match'] == 1)\n",
    "                                  & (current_match['s2v_score'] > threshold)]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).index[0]\n",
    "        inchikey_match = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_match == inchikey_query)\n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_s2v.append(best_bet)\n",
    "\n",
    "    #RF all data\n",
    "    predictions = all_result[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_rf_all.append(best_bet)\n",
    "    \n",
    "    #RF with top20\n",
    "    predictions = results_top20[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_rf_top20.append(best_bet)\n",
    "    \n",
    "    #RF with top20 50/50 split\n",
    "    predictions = results_top20_5050[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_rf_top20_5050.append(best_bet)\n",
    "\n",
    "\n",
    "\n",
    "# Make arrays from lists:\n",
    "test_m_s2v_arr = np.array(test_m_s2v)\n",
    "test_m_rf_all_arr = np.array(test_m_rf_all)\n",
    "test_m_rf_top20_arr = np.array(test_m_rf_top20)\n",
    "test_m_rf_top20_5050_arr = np.array(test_m_rf_top20_5050)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Spectral library matching comparison (>1 matching inchikey in library)')"
      ]
     },
     "execution_count": 132,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.style.use('ggplot')\n",
    "\n",
    "ylabels = ['mass + Spec2Vec',\n",
    "          'rf all',\n",
    "          'rf top20',\n",
    "          'rf top20 50/50-split']\n",
    "xlabels = ['true matches', 'false matches', 'no matches found']\n",
    "          #'mass + mod.cosine \\n + as backup: Spec2Vec + mod.cosine']\n",
    "data = np.array([[np.sum(test_m_s2v_arr == 1), np.sum(test_m_s2v_arr == 0), np.sum(test_m_s2v_arr == -1)],\n",
    "                [np.sum(test_m_rf_all_arr == 1), np.sum(test_m_rf_all_arr == 0), np.sum(test_m_rf_all_arr == -1)],\n",
    "               [np.sum(test_m_rf_top20_arr == 1), np.sum(test_m_rf_top20_arr == 0), np.sum(test_m_rf_top20_arr == -1)],\n",
    "               [np.sum(test_m_rf_top20_5050_arr == 1), np.sum(test_m_rf_top20_5050_arr == 0), np.sum(test_m_rf_top20_5050_arr == -1)]])\n",
    "  \n",
    "dataframe = pd.DataFrame(data, columns=xlabels, index=ylabels)\n",
    "\n",
    "ax = dataframe.sort_values(by=['false matches'], ascending=True).plot.barh(stacked=True, edgecolor='none')\n",
    "\n",
    "horiz_offset = 1.03\n",
    "vert_offset = 1.\n",
    "ax.legend(bbox_to_anchor=(horiz_offset, vert_offset))\n",
    "plt.xlabel('Number of spectra')\n",
    "plt.title('Spectral library matching comparison (>1 matching inchikey in library)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Calculate mass_match again:\n",
    "    -in bins of 2, 5\n",
    "    -in a percent change: abs((query-library)/library)*100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {},
   "outputs": [],
   "source": [
    "from rdkit import Chem, DataStructs\n",
    "\n",
    "def find_info_matches(matches, documents_library, documents_query,\n",
    "                      bins = [2, 5], calc_change = False, add_label = True,\n",
    "                      add_mass_match = True, add_tanimoto_sim = True, add_cols = ['parent_mass']):\n",
    "    '''\n",
    "    To each df in matches, add label for matching inchikey and mass matches\n",
    "    \n",
    "    A matching inchikey gets label 1 and non-matching 0. Mass matches are represented in discrete \n",
    "    bins or as a percentage of mass change.\n",
    "\n",
    "    matches: list of pandas DataFrames, library matching result of query on library\n",
    "    documents_library: list of SpectrumDocuments, spectra in library\n",
    "    documents_query: list of SpectrumDocuments, spectra in query set.\n",
    "        Indices should correspond to indices of matches.\n",
    "    bins: list of int/float, the cutoff for the mass change, will result in boolean columns,\n",
    "        i.e. if a cutoff is 2 there will be a column with 1 (true) or 0 (false) for if the mass change\n",
    "        is within 2\n",
    "    calc_change: bool, instead of bins add a percentage change of the parent mass difference\n",
    "    add_label: bool, add a label for inchikey match or not\n",
    "    add_mass_match: bool, add mass_match info or not\n",
    "    add_tanimoto_sim: bool, add tanimoto similarity or not\n",
    "    add_cols: bool/list of str, add other info present in metadata such as parent_mass, adduct\n",
    "        default: ['parent_mass']\n",
    "    \n",
    "    Output\n",
    "    matches_with_info: list of pandas DataFrames, library matching result of query\n",
    "        on library with matching labels\n",
    "    '''\n",
    "    matches_with_info = []\n",
    "    for query_id in range(len(documents_query)):\n",
    "        match = matches[query_id]\n",
    "        if add_label:\n",
    "            query_inchi = documents_query[query_id]._obj.get(\"inchikey\")[:14]\n",
    "            match = find_inchikey_match(match, documents_library, query_inchi)\n",
    "        if add_mass_match:\n",
    "            query_mass = documents_query[query_id]._obj.get(\"parent_mass\")\n",
    "            match = find_mass_matches(match, documents_library, query_mass, bins, calc_change)\n",
    "        if add_cols:\n",
    "            match = find_basic_info(match, documents_library, add_cols)\n",
    "        if add_tanimoto_sim:\n",
    "            query_smiles = documents_query[query_id]._obj.get(\"smiles\")\n",
    "            match = find_tanimoto_sim(match, documents_library, query_smiles)\n",
    "        matches_with_info.append(match)\n",
    "    return matches_with_info\n",
    "\n",
    "def find_inchikey_match(matches, documents_library, query_inchi):\n",
    "    '''To each match in the matches df, add label for matching inchikey (1) or non-matching (0)\n",
    "\n",
    "    matches: pandas DataFrame, library matching result of 1 query on library\n",
    "    documents_library: list of SpectrumDocuments, spectra in library\n",
    "    query_inchi: str, first 14 symbols of query inchikey\n",
    "    df: pandas DataFrame, library matching result of query on library with matching labels\n",
    "    '''\n",
    "    matches_with_labels = []\n",
    "    df = matches.copy()\n",
    "    library_ids = df.index.values\n",
    "    labels = []\n",
    "    for lib_id in library_ids:\n",
    "        lib_inchi = documents_library[lib_id]._obj.get(\"inchikey\")[:14]\n",
    "        lab = 0\n",
    "        if query_inchi == lib_inchi:\n",
    "            lab = 1\n",
    "        labels.append(lab)\n",
    "    df['label'] = labels\n",
    "    return df\n",
    "\n",
    "def find_mass_matches(matches, documents_library, query_mass, bins = [2, 5], calc_change = False):\n",
    "    '''\n",
    "    To each match in matches df, add the mass change of query to match in bins or a percentage change\n",
    "\n",
    "    matches: pandas DataFrame, library matching result of 1 query on library\n",
    "    documents_library: list of SpectrumDocuments, spectra in library\n",
    "    query_mass: float, parent mass of query\n",
    "    bins: list of int/float, the cutoff for the mass change, will result in boolean columns,\n",
    "        i.e. if a cutoff is 2 there will be a column with 1 (true) or 0 (false) for if the mass change\n",
    "        is within 2\n",
    "    calc_change: bool, instead of bins add a percentage change of the parent mass difference\n",
    "    df: pandas DataFrame, library matching result of 1 query on library with mass matches\n",
    "    '''\n",
    "    range_bins = range(len(bins)) #calc once\n",
    "    df = matches.copy()\n",
    "    library_ids = df.index.values\n",
    "    masses = [[] for _ in range_bins] #initialise\n",
    "    mass_changes = []\n",
    "    for lib_id in library_ids:\n",
    "        lib_mass = documents_library[lib_id]._obj.get(\"parent_mass\")\n",
    "        if calc_change: #calculate a percentage change in parent mass instead of discrete bins\n",
    "            perc_change = abs(lib_mass - query_mass) / query_mass * 100\n",
    "            mass_changes.append(perc_change)\n",
    "        else:\n",
    "            for bin_i in range_bins:\n",
    "                cutoff = bins[bin_i]\n",
    "                lab = 0\n",
    "                if abs(query_mass - lib_mass) < cutoff:\n",
    "                    lab = 1\n",
    "                masses[bin_i].append(lab)\n",
    "    #add to df\n",
    "    if calc_change:\n",
    "        df['perc_mass_change'] = mass_changes\n",
    "    else:\n",
    "        for bin_i in range_bins:\n",
    "            df['mass_match_' + str(bins[bin_i])] = masses[bin_i]\n",
    "    return df\n",
    "\n",
    "def find_basic_info(matches, documents_library, add_cols = ['parent_mass']):\n",
    "    '''\n",
    "    To each match in matches df, add the info from add_cols entries\n",
    "\n",
    "    matches: pandas DataFrame, library matching result of 1 query on library\n",
    "    documents_library: list of SpectrumDocuments, spectra in library\n",
    "    df: pandas DataFrame, library matching result of 1 query on library with added info\n",
    "    '''\n",
    "    df = matches.copy()\n",
    "    library_ids = df.index.values\n",
    "    if add_cols:\n",
    "        for col in add_cols:\n",
    "            col_data = []\n",
    "            for lib_id in library_ids:\n",
    "                lib_data = documents_library[lib_id]._obj.get(col)\n",
    "                col_data.append(lib_data)\n",
    "            df[col] = col_data\n",
    "    return df\n",
    "\n",
    "def find_tanimoto_sim(matches, documents_library, query_smiles):\n",
    "    '''To each match in matches df, add the tanimoto similarity between query and match\n",
    "\n",
    "    matches: pandas DataFrame, library matching result of 1 query on library\n",
    "    documents_library: list of SpectrumDocuments, spectra in library\n",
    "    df: pandas DataFrame, library matching result of 1 query on library with tanimoto similarities\n",
    "    '''\n",
    "    df = matches.copy()\n",
    "    sims = []\n",
    "    library_ids = df.index.values\n",
    "    \n",
    "    if not query_smiles or query_smiles == \"None\": # check that query smiles exist\n",
    "        df['similarity'] = [0] * len(library_ids) #default to all 0 if it doesnt exist\n",
    "        return df\n",
    "    ms_q = Chem.MolFromSmiles(query_smiles)\n",
    "    if not ms_q: #in case something is wrong with smiles\n",
    "        df['similarity'] = [0] * len(library_ids) #default to all 0 if it doesnt exist\n",
    "        return df\n",
    "    \n",
    "    fp_q = Chem.RDKFingerprint(ms_q)\n",
    "    for lib_id in library_ids:\n",
    "        smiles_lib = documents_library[lib_id]._obj.get(\"smiles\")\n",
    "        if smiles_lib and smiles_lib != \"None\":\n",
    "            ms_lib = Chem.MolFromSmiles(smiles_lib)\n",
    "            if ms_lib:\n",
    "                fp_lib = Chem.RDKFingerprint(ms_lib)\n",
    "                score = DataStructs.FingerprintSimilarity(fp_q, fp_lib)\n",
    "            else: #in case something is wrong with smiles\n",
    "                score = 0\n",
    "        else: #in case it doesnt have smiles\n",
    "            score = 0\n",
    "        sims.append(score)\n",
    "    df['similarity'] = sims\n",
    "    return df\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>label</th>\n",
       "      <th>mass_match_2</th>\n",
       "      <th>mass_match_5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>26647</th>\n",
       "      <td>0.998302</td>\n",
       "      <td>13</td>\n",
       "      <td>0.998302</td>\n",
       "      <td>13</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.738719</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1652</th>\n",
       "      <td>0.005188</td>\n",
       "      <td>6</td>\n",
       "      <td>0.885819</td>\n",
       "      <td>15</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.607773</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27002</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.387876</td>\n",
       "      <td>12</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.595645</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2166</th>\n",
       "      <td>0.006927</td>\n",
       "      <td>5</td>\n",
       "      <td>0.011513</td>\n",
       "      <td>7</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.578534</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1636</th>\n",
       "      <td>0.005002</td>\n",
       "      <td>8</td>\n",
       "      <td>0.913320</td>\n",
       "      <td>9</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.562214</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1643</th>\n",
       "      <td>0.002817</td>\n",
       "      <td>4</td>\n",
       "      <td>0.902918</td>\n",
       "      <td>10</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.552722</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1649</th>\n",
       "      <td>0.158497</td>\n",
       "      <td>5</td>\n",
       "      <td>0.734370</td>\n",
       "      <td>8</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.543317</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1647</th>\n",
       "      <td>0.000015</td>\n",
       "      <td>1</td>\n",
       "      <td>0.903097</td>\n",
       "      <td>9</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.540782</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27181</th>\n",
       "      <td>0.000132</td>\n",
       "      <td>3</td>\n",
       "      <td>0.000153</td>\n",
       "      <td>5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.529925</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27309</th>\n",
       "      <td>0.000135</td>\n",
       "      <td>1</td>\n",
       "      <td>0.025023</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.528484</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1631</th>\n",
       "      <td>0.000209</td>\n",
       "      <td>3</td>\n",
       "      <td>0.902313</td>\n",
       "      <td>10</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.523983</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27012</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.846120</td>\n",
       "      <td>11</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.519714</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13072</th>\n",
       "      <td>0.000025</td>\n",
       "      <td>4</td>\n",
       "      <td>0.001060</td>\n",
       "      <td>6</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.515517</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27469</th>\n",
       "      <td>0.000542</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000542</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.514389</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27368</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000143</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.512785</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2825</th>\n",
       "      <td>0.000024</td>\n",
       "      <td>3</td>\n",
       "      <td>0.001772</td>\n",
       "      <td>5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.512240</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1193</th>\n",
       "      <td>0.000042</td>\n",
       "      <td>3</td>\n",
       "      <td>0.000042</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.509906</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27340</th>\n",
       "      <td>0.000773</td>\n",
       "      <td>6</td>\n",
       "      <td>0.010043</td>\n",
       "      <td>14</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.502886</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27466</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000051</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.500744</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4602</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.001828</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.499288</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "26647      0.998302              13          0.998302                  13   \n",
       "1652       0.005188               6          0.885819                  15   \n",
       "27002      0.000000               0          0.387876                  12   \n",
       "2166       0.006927               5          0.011513                   7   \n",
       "1636       0.005002               8          0.913320                   9   \n",
       "1643       0.002817               4          0.902918                  10   \n",
       "1649       0.158497               5          0.734370                   8   \n",
       "1647       0.000015               1          0.903097                   9   \n",
       "27181      0.000132               3          0.000153                   5   \n",
       "27309      0.000135               1          0.025023                   2   \n",
       "1631       0.000209               3          0.902313                  10   \n",
       "27012      0.000000               0          0.846120                  11   \n",
       "13072      0.000025               4          0.001060                   6   \n",
       "27469      0.000542               1          0.000542                   1   \n",
       "27368      0.000000               0          0.000143                   2   \n",
       "2825       0.000024               3          0.001772                   5   \n",
       "1193       0.000042               3          0.000042                   3   \n",
       "27340      0.000773               6          0.010043                  14   \n",
       "27466      0.000000               0          0.000051                   1   \n",
       "4602       0.000000               0          0.001828                   2   \n",
       "\n",
       "       mass_match  s2v_score  label  mass_match_2  mass_match_5  \n",
       "26647         0.0   0.738719      1             0             0  \n",
       "1652          0.0   0.607773      0             0             0  \n",
       "27002         0.0   0.595645      0             0             0  \n",
       "2166          0.0   0.578534      0             0             0  \n",
       "1636          0.0   0.562214      0             0             0  \n",
       "1643          0.0   0.552722      0             0             0  \n",
       "1649          0.0   0.543317      0             0             0  \n",
       "1647          0.0   0.540782      0             0             0  \n",
       "27181         0.0   0.529925      0             0             0  \n",
       "27309         0.0   0.528484      0             0             0  \n",
       "1631          0.0   0.523983      0             0             0  \n",
       "27012         0.0   0.519714      0             0             0  \n",
       "13072         0.0   0.515517      0             0             0  \n",
       "27469         0.0   0.514389      0             0             0  \n",
       "27368         0.0   0.512785      0             0             0  \n",
       "2825          0.0   0.512240      0             0             0  \n",
       "1193          0.0   0.509906      0             0             0  \n",
       "27340         0.0   0.502886      0             0             0  \n",
       "27466         0.0   0.500744      0             0             0  \n",
       "4602          0.0   0.499288      0             0             0  "
      ]
     },
     "execution_count": 134,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "found_matches_s2v_more_info = find_info_matches(found_matches_s2v, documents_library_s2v, documents_query_s2v, bins = [2,5],\n",
    "                                                add_cols=False, add_tanimoto_sim=False)\n",
    "found_matches_s2v_mm5_pmass_sim = find_info_matches(found_matches_s2v, documents_library_s2v, documents_query_s2v, bins = [5])\n",
    "found_matches_s2v_more_info[171].sort_values(by=['s2v_score'], ascending=False).iloc[:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(239233, 9) 239233\n",
      "Label distribution: {0: 228606, 1: 10627}\n",
      "Training Features Shape: (179424, 8)\n",
      "Training Labels Shape: (179424,)\n",
      "Training label distribution: {0: 171430, 1: 7994}\n",
      "Testing Features Shape: (59809, 8)\n",
      "Testing Labels Shape: (59809,)\n",
      "Test label distribution: {0: 57176, 1: 2633}\n",
      "\n",
      "Predicted label distribution: {0: 57876, 1: 1933}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "56285 1042\n",
      "0->1 1->0\n",
      "891 1591\n",
      "Accuracy: 0.96\n",
      "\n",
      "Variable: s2v_score            Importance: 0.45\n",
      "Variable: cosine_score         Importance: 0.19\n",
      "Variable: mod_cosine_score     Importance: 0.17\n",
      "Variable: cosine_matches       Importance: 0.07\n",
      "Variable: mod_cosine_matches   Importance: 0.07\n",
      "Variable: mass_match           Importance: 0.03\n",
      "Variable: mass_match_2         Importance: 0.01\n",
      "Variable: mass_match_5         Importance: 0.01\n"
     ]
    }
   ],
   "source": [
    "#RF with two mass bins extra <2 dalton difference and < 5 dalton difference in parent mass\n",
    "all_matches_s2v_more_info = found_matches_s2v_more_info[0].append(found_matches_s2v_more_info[1:])\n",
    "print(all_matches_s2v_more_info.shape, sum(df.shape[0] for df in found_matches_s2v_more_info))\n",
    "rf_all_more_info = do_random_forest(all_matches_s2v_more_info, 'label', verbose = True, num_est = 200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label distribution: {0: 228606, 1: 10627}\n",
      "Training Features Shape: (179424, 8)\n",
      "Training Labels Shape: (179424,)\n",
      "Training label distribution: {0: 171430, 1: 7994}\n",
      "Testing Features Shape: (59809, 8)\n",
      "Testing Labels Shape: (59809,)\n",
      "Test label distribution: {0: 57176, 1: 2633}\n",
      "\n",
      "Predicted label distribution: {0: 58012, 1: 1797}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "56816 1437\n",
      "0->1 1->0\n",
      "360 1196\n",
      "Accuracy: 0.97\n",
      "\n",
      "Variable: s2v_score            Importance: 0.31\n",
      "Variable: parent_mass          Importance: 0.24\n",
      "Variable: cosine_score         Importance: 0.14\n",
      "Variable: mod_cosine_score     Importance: 0.12\n",
      "Variable: cosine_matches       Importance: 0.06\n",
      "Variable: mod_cosine_matches   Importance: 0.06\n",
      "Variable: mass_match           Importance: 0.04\n",
      "Variable: mass_match_5         Importance: 0.02\n"
     ]
    }
   ],
   "source": [
    "#RF with parent mass\n",
    "all_matches_s2v_mm5_pmass = found_matches_s2v_mm5_pmass_sim[0].append(found_matches_s2v_mm5_pmass_sim[1:]).drop('similarity', axis=1)\n",
    "all_result_mm5_pmass = do_random_forest(all_matches_s2v_mm5_pmass, 'label', verbose = True, num_est = 200)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Random forest regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label distribution: {0: 228606, 1: 10627}\n",
      "Training Features Shape: (179424, 7)\n",
      "Training Labels Shape: (179424,)\n",
      "Training label distribution: {0: 171430, 1: 7994}\n",
      "Testing Features Shape: (59809, 7)\n",
      "Testing Labels Shape: (59809,)\n",
      "Test label distribution: {0: 57176, 1: 2633}\n",
      "\n",
      "Predicted label distribution: {0.0: 54269, 0.1: 1298, 0.2: 727, 0.3: 532, 0.4: 443, 0.5: 309, 0.6: 369, 0.7: 440, 0.8: 415, 0.9: 493, 1.0: 514}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "56516 1719\n",
      "0->1 1->0\n",
      "660 914\n",
      "Accuracy: 0.97\n",
      "\n",
      "Variable: perc_mass_change     Importance: 0.5\n",
      "Variable: s2v_score            Importance: 0.25\n",
      "Variable: cosine_score         Importance: 0.08\n",
      "Variable: mod_cosine_score     Importance: 0.08\n",
      "Variable: cosine_matches       Importance: 0.04\n",
      "Variable: mod_cosine_matches   Importance: 0.04\n",
      "Variable: mass_match           Importance: 0.0\n"
     ]
    }
   ],
   "source": [
    "#with mass change perc\n",
    "found_matches_s2v_match_perc = find_info_matches(found_matches_s2v, documents_library_s2v, documents_query_s2v, calc_change=True, add_tanimoto_sim=False, add_cols=False)\n",
    "#found_matches_s2v_match_perc[171].sort_values(by=['s2v_score'], ascending=False).iloc[:20]\n",
    "all_matches_s2v_match_perc = found_matches_s2v_match_perc[0].append(found_matches_s2v_match_perc[1:])\n",
    "rfreg_all_match_perc = do_random_forest(all_matches_s2v_match_perc, 'label', rf_regr = True,  verbose = True, num_est = 200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>label</th>\n",
       "      <th>perc_mass_change</th>\n",
       "      <th>prediction</th>\n",
       "      <th>correct_guess</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>26647</th>\n",
       "      <td>0.998302</td>\n",
       "      <td>13</td>\n",
       "      <td>0.998302</td>\n",
       "      <td>13</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.738719</td>\n",
       "      <td>1</td>\n",
       "      <td>96.282100</td>\n",
       "      <td>0.555</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1652</th>\n",
       "      <td>0.005188</td>\n",
       "      <td>6</td>\n",
       "      <td>0.885819</td>\n",
       "      <td>15</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.607773</td>\n",
       "      <td>0</td>\n",
       "      <td>47.710866</td>\n",
       "      <td>0.010</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27002</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.387876</td>\n",
       "      <td>12</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.595645</td>\n",
       "      <td>0</td>\n",
       "      <td>0.846164</td>\n",
       "      <td>0.040</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2166</th>\n",
       "      <td>0.006927</td>\n",
       "      <td>5</td>\n",
       "      <td>0.011513</td>\n",
       "      <td>7</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.578534</td>\n",
       "      <td>0</td>\n",
       "      <td>35.868966</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1636</th>\n",
       "      <td>0.005002</td>\n",
       "      <td>8</td>\n",
       "      <td>0.913320</td>\n",
       "      <td>9</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.562214</td>\n",
       "      <td>0</td>\n",
       "      <td>42.636756</td>\n",
       "      <td>0.005</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1643</th>\n",
       "      <td>0.002817</td>\n",
       "      <td>4</td>\n",
       "      <td>0.902918</td>\n",
       "      <td>10</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.552722</td>\n",
       "      <td>0</td>\n",
       "      <td>42.636925</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1649</th>\n",
       "      <td>0.158497</td>\n",
       "      <td>5</td>\n",
       "      <td>0.734370</td>\n",
       "      <td>8</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.543317</td>\n",
       "      <td>0</td>\n",
       "      <td>37.561971</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1647</th>\n",
       "      <td>0.000015</td>\n",
       "      <td>1</td>\n",
       "      <td>0.903097</td>\n",
       "      <td>9</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.540782</td>\n",
       "      <td>0</td>\n",
       "      <td>47.371149</td>\n",
       "      <td>0.010</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27181</th>\n",
       "      <td>0.000132</td>\n",
       "      <td>3</td>\n",
       "      <td>0.000153</td>\n",
       "      <td>5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.529925</td>\n",
       "      <td>0</td>\n",
       "      <td>190.684295</td>\n",
       "      <td>0.005</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27309</th>\n",
       "      <td>0.000135</td>\n",
       "      <td>1</td>\n",
       "      <td>0.025023</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.528484</td>\n",
       "      <td>0</td>\n",
       "      <td>178.165667</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1631</th>\n",
       "      <td>0.000209</td>\n",
       "      <td>3</td>\n",
       "      <td>0.902313</td>\n",
       "      <td>10</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.523983</td>\n",
       "      <td>0</td>\n",
       "      <td>42.295518</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27012</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.846120</td>\n",
       "      <td>11</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.519714</td>\n",
       "      <td>0</td>\n",
       "      <td>3.550879</td>\n",
       "      <td>0.005</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13072</th>\n",
       "      <td>0.000025</td>\n",
       "      <td>4</td>\n",
       "      <td>0.001060</td>\n",
       "      <td>6</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.515517</td>\n",
       "      <td>0</td>\n",
       "      <td>24.698384</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27469</th>\n",
       "      <td>0.000542</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000542</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.514389</td>\n",
       "      <td>0</td>\n",
       "      <td>95.088271</td>\n",
       "      <td>0.005</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27368</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000143</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.512785</td>\n",
       "      <td>0</td>\n",
       "      <td>190.349144</td>\n",
       "      <td>0.005</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "26647      0.998302              13          0.998302                  13   \n",
       "1652       0.005188               6          0.885819                  15   \n",
       "27002      0.000000               0          0.387876                  12   \n",
       "2166       0.006927               5          0.011513                   7   \n",
       "1636       0.005002               8          0.913320                   9   \n",
       "1643       0.002817               4          0.902918                  10   \n",
       "1649       0.158497               5          0.734370                   8   \n",
       "1647       0.000015               1          0.903097                   9   \n",
       "27181      0.000132               3          0.000153                   5   \n",
       "27309      0.000135               1          0.025023                   2   \n",
       "1631       0.000209               3          0.902313                  10   \n",
       "27012      0.000000               0          0.846120                  11   \n",
       "13072      0.000025               4          0.001060                   6   \n",
       "27469      0.000542               1          0.000542                   1   \n",
       "27368      0.000000               0          0.000143                   2   \n",
       "\n",
       "       mass_match  s2v_score  label  perc_mass_change  prediction  \\\n",
       "26647         0.0   0.738719      1         96.282100       0.555   \n",
       "1652          0.0   0.607773      0         47.710866       0.010   \n",
       "27002         0.0   0.595645      0          0.846164       0.040   \n",
       "2166          0.0   0.578534      0         35.868966       0.000   \n",
       "1636          0.0   0.562214      0         42.636756       0.005   \n",
       "1643          0.0   0.552722      0         42.636925       0.000   \n",
       "1649          0.0   0.543317      0         37.561971       0.000   \n",
       "1647          0.0   0.540782      0         47.371149       0.010   \n",
       "27181         0.0   0.529925      0        190.684295       0.005   \n",
       "27309         0.0   0.528484      0        178.165667       0.000   \n",
       "1631          0.0   0.523983      0         42.295518       0.000   \n",
       "27012         0.0   0.519714      0          3.550879       0.005   \n",
       "13072         0.0   0.515517      0         24.698384       0.000   \n",
       "27469         0.0   0.514389      0         95.088271       0.005   \n",
       "27368         0.0   0.512785      0        190.349144       0.005   \n",
       "\n",
       "       correct_guess  \n",
       "26647           True  \n",
       "1652            True  \n",
       "27002           True  \n",
       "2166            True  \n",
       "1636            True  \n",
       "1643            True  \n",
       "1649            True  \n",
       "1647            True  \n",
       "27181           True  \n",
       "27309           True  \n",
       "1631            True  \n",
       "27012           True  \n",
       "13072           True  \n",
       "27469           True  \n",
       "27368           True  "
      ]
     },
     "execution_count": 138,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_select = found_matches_s2v_match_perc[171].copy()\n",
    "dropped_df_select = df_select.drop(['label'],axis=1)\n",
    "predictions = rfreg_all_match_perc[3].predict(dropped_df_select)\n",
    "df_select['prediction'] = predictions\n",
    "df_select['correct_guess'] = (round(df_select['label']) == round(df_select['prediction']))\n",
    "df_select.sort_values(by=['s2v_score'], ascending=False).iloc[:15]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(59809, 7)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>perc_mass_change</th>\n",
       "      <th>label</th>\n",
       "      <th>prediction</th>\n",
       "      <th>correct_guess</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>35122</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>15.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>15.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34791</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>41.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>41.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>0.960</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40735</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>123.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>123.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>0.510</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1733</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>16.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>16.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40191</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>55.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>55.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>0.560</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25388</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>49.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>49.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>0.990</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44288</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>34.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>34.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>0.935</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26903</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>24.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>24.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49212</th>\n",
       "      <td>0.999343</td>\n",
       "      <td>16.0</td>\n",
       "      <td>0.999343</td>\n",
       "      <td>16.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.999699</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25597</th>\n",
       "      <td>0.999543</td>\n",
       "      <td>13.0</td>\n",
       "      <td>0.999543</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.998715</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>0.940</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16448</th>\n",
       "      <td>0.999735</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.999753</td>\n",
       "      <td>68.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.998407</td>\n",
       "      <td>0.246150</td>\n",
       "      <td>0</td>\n",
       "      <td>0.005</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32228</th>\n",
       "      <td>0.999806</td>\n",
       "      <td>61.0</td>\n",
       "      <td>0.999808</td>\n",
       "      <td>62.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.998219</td>\n",
       "      <td>7.154190</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36591</th>\n",
       "      <td>0.999870</td>\n",
       "      <td>70.0</td>\n",
       "      <td>0.999896</td>\n",
       "      <td>71.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.998122</td>\n",
       "      <td>6.507486</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4482</th>\n",
       "      <td>0.999829</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.999829</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997947</td>\n",
       "      <td>1.110912</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10890</th>\n",
       "      <td>0.999955</td>\n",
       "      <td>72.0</td>\n",
       "      <td>0.999972</td>\n",
       "      <td>73.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997897</td>\n",
       "      <td>0.262298</td>\n",
       "      <td>0</td>\n",
       "      <td>0.010</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31930</th>\n",
       "      <td>0.999890</td>\n",
       "      <td>66.0</td>\n",
       "      <td>0.999892</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997840</td>\n",
       "      <td>7.411323</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21224</th>\n",
       "      <td>0.999842</td>\n",
       "      <td>59.0</td>\n",
       "      <td>0.999842</td>\n",
       "      <td>59.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997809</td>\n",
       "      <td>3.699863</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>53364</th>\n",
       "      <td>0.999891</td>\n",
       "      <td>61.0</td>\n",
       "      <td>0.999891</td>\n",
       "      <td>61.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.997776</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>0.870</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8183</th>\n",
       "      <td>0.999908</td>\n",
       "      <td>74.0</td>\n",
       "      <td>0.999908</td>\n",
       "      <td>74.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997749</td>\n",
       "      <td>10.544870</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27328</th>\n",
       "      <td>0.999741</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.999741</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997732</td>\n",
       "      <td>1.110912</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41902</th>\n",
       "      <td>0.999559</td>\n",
       "      <td>64.0</td>\n",
       "      <td>0.999578</td>\n",
       "      <td>65.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997677</td>\n",
       "      <td>10.758157</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12282</th>\n",
       "      <td>0.999770</td>\n",
       "      <td>72.0</td>\n",
       "      <td>0.999782</td>\n",
       "      <td>73.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997638</td>\n",
       "      <td>3.122594</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11910</th>\n",
       "      <td>0.999868</td>\n",
       "      <td>63.0</td>\n",
       "      <td>0.999868</td>\n",
       "      <td>63.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.997598</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>0.860</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24050</th>\n",
       "      <td>0.999956</td>\n",
       "      <td>61.0</td>\n",
       "      <td>0.999956</td>\n",
       "      <td>61.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997564</td>\n",
       "      <td>7.449731</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>53201</th>\n",
       "      <td>0.999852</td>\n",
       "      <td>18.0</td>\n",
       "      <td>0.999852</td>\n",
       "      <td>18.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.997526</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.000</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29459</th>\n",
       "      <td>0.999944</td>\n",
       "      <td>65.0</td>\n",
       "      <td>0.999957</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997498</td>\n",
       "      <td>0.271366</td>\n",
       "      <td>0</td>\n",
       "      <td>0.010</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17655</th>\n",
       "      <td>0.999744</td>\n",
       "      <td>63.0</td>\n",
       "      <td>0.999744</td>\n",
       "      <td>63.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997466</td>\n",
       "      <td>0.847415</td>\n",
       "      <td>0</td>\n",
       "      <td>0.050</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>983</th>\n",
       "      <td>0.999904</td>\n",
       "      <td>58.0</td>\n",
       "      <td>0.999904</td>\n",
       "      <td>58.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.997382</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.635</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38689</th>\n",
       "      <td>0.999719</td>\n",
       "      <td>62.0</td>\n",
       "      <td>0.999728</td>\n",
       "      <td>64.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997378</td>\n",
       "      <td>3.319897</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27847</th>\n",
       "      <td>0.999834</td>\n",
       "      <td>58.0</td>\n",
       "      <td>0.999834</td>\n",
       "      <td>58.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997356</td>\n",
       "      <td>1.181723</td>\n",
       "      <td>0</td>\n",
       "      <td>0.010</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "35122      1.000000            15.0          1.000000                15.0   \n",
       "34791      1.000000            41.0          1.000000                41.0   \n",
       "40735      1.000000           123.0          1.000000               123.0   \n",
       "1733       1.000000            16.0          1.000000                16.0   \n",
       "40191      1.000000            55.0          1.000000                55.0   \n",
       "25388      1.000000            49.0          1.000000                49.0   \n",
       "44288      1.000000            34.0          1.000000                34.0   \n",
       "26903      1.000000            24.0          1.000000                24.0   \n",
       "49212      0.999343            16.0          0.999343                16.0   \n",
       "25597      0.999543            13.0          0.999543                13.0   \n",
       "16448      0.999735            67.0          0.999753                68.0   \n",
       "32228      0.999806            61.0          0.999808                62.0   \n",
       "36591      0.999870            70.0          0.999896                71.0   \n",
       "4482       0.999829            67.0          0.999829                67.0   \n",
       "10890      0.999955            72.0          0.999972                73.0   \n",
       "31930      0.999890            66.0          0.999892                67.0   \n",
       "21224      0.999842            59.0          0.999842                59.0   \n",
       "53364      0.999891            61.0          0.999891                61.0   \n",
       "8183       0.999908            74.0          0.999908                74.0   \n",
       "27328      0.999741            67.0          0.999741                67.0   \n",
       "41902      0.999559            64.0          0.999578                65.0   \n",
       "12282      0.999770            72.0          0.999782                73.0   \n",
       "11910      0.999868            63.0          0.999868                63.0   \n",
       "24050      0.999956            61.0          0.999956                61.0   \n",
       "53201      0.999852            18.0          0.999852                18.0   \n",
       "29459      0.999944            65.0          0.999957                67.0   \n",
       "17655      0.999744            63.0          0.999744                63.0   \n",
       "983        0.999904            58.0          0.999904                58.0   \n",
       "38689      0.999719            62.0          0.999728                64.0   \n",
       "27847      0.999834            58.0          0.999834                58.0   \n",
       "\n",
       "       mass_match  s2v_score  perc_mass_change  label  prediction  \\\n",
       "35122         1.0   1.000000          0.000000      1       1.000   \n",
       "34791         1.0   1.000000          0.000000      1       0.960   \n",
       "40735         1.0   1.000000          0.000000      1       0.510   \n",
       "1733          1.0   1.000000          0.000000      1       1.000   \n",
       "40191         1.0   1.000000          0.000000      1       0.560   \n",
       "25388         1.0   1.000000          0.000000      1       0.990   \n",
       "44288         1.0   1.000000          0.000000      1       0.935   \n",
       "26903         1.0   1.000000          0.000000      1       1.000   \n",
       "49212         1.0   0.999699          0.000000      1       1.000   \n",
       "25597         1.0   0.998715          0.000000      1       0.940   \n",
       "16448         0.0   0.998407          0.246150      0       0.005   \n",
       "32228         0.0   0.998219          7.154190      0       0.000   \n",
       "36591         0.0   0.998122          6.507486      0       0.000   \n",
       "4482          0.0   0.997947          1.110912      0       0.000   \n",
       "10890         0.0   0.997897          0.262298      0       0.010   \n",
       "31930         0.0   0.997840          7.411323      0       0.000   \n",
       "21224         0.0   0.997809          3.699863      0       0.000   \n",
       "53364         1.0   0.997776          0.000000      1       0.870   \n",
       "8183          0.0   0.997749         10.544870      0       0.000   \n",
       "27328         0.0   0.997732          1.110912      0       0.000   \n",
       "41902         0.0   0.997677         10.758157      0       0.000   \n",
       "12282         0.0   0.997638          3.122594      0       0.000   \n",
       "11910         1.0   0.997598          0.000000      1       0.860   \n",
       "24050         0.0   0.997564          7.449731      0       0.000   \n",
       "53201         1.0   0.997526          0.000000      0       1.000   \n",
       "29459         0.0   0.997498          0.271366      0       0.010   \n",
       "17655         0.0   0.997466          0.847415      0       0.050   \n",
       "983           1.0   0.997382          0.000000      0       0.635   \n",
       "38689         0.0   0.997378          3.319897      0       0.000   \n",
       "27847         0.0   0.997356          1.181723      0       0.010   \n",
       "\n",
       "       correct_guess  \n",
       "35122           True  \n",
       "34791           True  \n",
       "40735           True  \n",
       "1733            True  \n",
       "40191           True  \n",
       "25388           True  \n",
       "44288           True  \n",
       "26903           True  \n",
       "49212           True  \n",
       "25597           True  \n",
       "16448           True  \n",
       "32228           True  \n",
       "36591           True  \n",
       "4482            True  \n",
       "10890           True  \n",
       "31930           True  \n",
       "21224           True  \n",
       "53364           True  \n",
       "8183            True  \n",
       "27328           True  \n",
       "41902           True  \n",
       "12282           True  \n",
       "11910           True  \n",
       "24050           True  \n",
       "53201          False  \n",
       "29459           True  \n",
       "17655           True  \n",
       "983            False  \n",
       "38689           True  \n",
       "27847           True  "
      ]
     },
     "execution_count": 139,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#visualising performance on the test set\n",
    "print(rfreg_all_match_perc[0].shape)\n",
    "test_set_rfreg_all_match_perc = pd.DataFrame(rfreg_all_match_perc[0], columns = cols+['perc_mass_change'])\n",
    "test_set_rfreg_all_match_perc['label'] = rfreg_all_match_perc[1]\n",
    "test_set_rfreg_all_match_perc['prediction'] = rfreg_all_match_perc[2]\n",
    "test_set_rfreg_all_match_perc['correct_guess'] = (round(test_set_rfreg_all_match_perc['label']) == round(test_set_rfreg_all_match_perc['prediction']))\n",
    "test_set_rfreg_all_match_perc.sort_values(by=['s2v_score'],ascending=False).iloc[:30]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label distribution: {0: 228606, 1: 10627}\n",
      "Training Features Shape: (179424, 8)\n",
      "Training Labels Shape: (179424,)\n",
      "Training label distribution: {0: 171430, 1: 7994}\n",
      "Testing Features Shape: (59809, 8)\n",
      "Testing Labels Shape: (59809,)\n",
      "Test label distribution: {0: 57176, 1: 2633}\n",
      "\n",
      "Predicted label distribution: {0.0: 54699, 0.1: 1248, 0.2: 635, 0.3: 420, 0.4: 306, 0.5: 242, 0.6: 262, 0.7: 249, 0.8: 285, 0.9: 403, 1.0: 1060}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "56807 2013\n",
      "0->1 1->0\n",
      "369 620\n",
      "Accuracy: 0.98\n",
      "\n",
      "Variable: perc_mass_change     Importance: 0.48\n",
      "Variable: parent_mass          Importance: 0.19\n",
      "Variable: s2v_score            Importance: 0.16\n",
      "Variable: cosine_score         Importance: 0.05\n",
      "Variable: mod_cosine_score     Importance: 0.05\n",
      "Variable: cosine_matches       Importance: 0.03\n",
      "Variable: mod_cosine_matches   Importance: 0.03\n",
      "Variable: mass_match           Importance: 0.0\n"
     ]
    }
   ],
   "source": [
    "#RF regression with parent mass and perc change\n",
    "found_matches_s2v_match_perc_pmass_sim = find_info_matches(found_matches_s2v, documents_library_s2v, documents_query_s2v, calc_change=True, add_cols = ['parent_mass'])\n",
    "found_matches_s2v_match_perc_pmass_sim[171].sort_values(by=['s2v_score'], ascending=False).iloc[:20]\n",
    "all_matches_s2v_match_perc_pmass = found_matches_s2v_match_perc_pmass_sim[0].append(found_matches_s2v_match_perc_pmass_sim[1:]).drop('similarity',axis=1)\n",
    "rfreg_all_match_perc_pmass = do_random_forest(all_matches_s2v_match_perc_pmass, 'label', rf_regr = True,  verbose = True, num_est = 200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C:\\Users\\joris\\Documents\\eScience_data\\data\\rfreg_all_match_perc_pmass.pickle\n",
      "C:\\Users\\joris\\Documents\\eScience_data\\data\\found_matches_s2v_match_perc_pmass_sim.pickle\n"
     ]
    }
   ],
   "source": [
    "import pickle\n",
    "outfile = os.path.join(path_data, 'rfreg_all_match_perc_pmass.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        rfreg_all_match_perc_pmass = pickle.load(inf)\n",
    "else:\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(rfreg_all_match_perc_pmass, outf)\n",
    "        \n",
    "        \n",
    "outfile = os.path.join(path_data, 'found_matches_s2v_match_perc_pmass_sim.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        found_matches_s2v_match_perc_pmass_sim = pickle.load(inf)\n",
    "else:\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(found_matches_s2v_match_perc_pmass_sim, outf)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Linear regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LinearRegression\n",
    "\n",
    "def do_linear_regression(feature_df, label_name, verbose = True):\n",
    "    '''Split data in training and test, train regressor, predict on test\n",
    "\n",
    "    feature_df: pandas DataFrame\n",
    "    label_name: str, name of the label column in feature_df\n",
    "    verbose: bool, print more info\n",
    "    \n",
    "    Returns test_features, test_labels, predictions, regressor\n",
    "    '''\n",
    "    #store labels\n",
    "    labels = np.array(feature_df[label_name])\n",
    "    #remove labels\n",
    "    feature_df_nolab = feature_df.drop(label_name, axis = 1)\n",
    "    feature_list = list(feature_df_nolab.columns)\n",
    "    feature_df_nolab = np.array(feature_df_nolab)\n",
    "    \n",
    "    #split data in training and test set\n",
    "    train_features, test_features,\\\n",
    "    train_labels, test_labels = train_test_split(feature_df_nolab, labels,\n",
    "                                                 test_size = 0.25, random_state = 42)\n",
    "    if verbose:\n",
    "        print('Label distribution:', dict(zip(*np.unique(np.round(labels,1), return_counts=True))))\n",
    "        print('Training Features Shape:', train_features.shape)\n",
    "        print('Training Labels Shape:', train_labels.shape)\n",
    "        print('Training label distribution:', dict(zip(*np.unique(np.round(train_labels,1), return_counts=True))))\n",
    "        print('Testing Features Shape:', test_features.shape)\n",
    "        print('Testing Labels Shape:', test_labels.shape)\n",
    "        print('Test label distribution:', dict(zip(*np.unique(np.round(test_labels,1), return_counts=True))))\n",
    "    \n",
    "    #train model\n",
    "    regressor = LinearRegression()\n",
    "    regressor.fit(train_features, train_labels)\n",
    "    \n",
    "    #predict test set\n",
    "    predictions = regressor.predict(test_features)\n",
    "\n",
    "    print('\\nPredicted label distribution:', dict(zip(*np.unique(np.round(predictions,1), return_counts=True))))\n",
    "    df = pd.DataFrame({'Actual': test_labels, 'Predicted': predictions})\n",
    "    print('Predictions:','\\n0->0','1->1')\n",
    "    zero_to_zero = np.sum((df.Actual < 0.5) & (df.Predicted < 0.5))\n",
    "    one_to_one = np.sum((df.Actual >= 0.5) & (df.Predicted >= 0.5))\n",
    "    print(zero_to_zero, one_to_one)\n",
    "    print('0->1','1->0')\n",
    "    print(np.sum((df.Actual < 0.5) & (df.Predicted >= 0.5)), \n",
    "    np.sum((df.Actual >= 0.5) & (df.Predicted < 0.5)))\n",
    "    print(\"Accuracy: {:.2f}\\n\".format((zero_to_zero + one_to_one)/len(predictions)))\n",
    "\n",
    "    #get feature importances\n",
    "    importances = list(regressor.coef_)\n",
    "    #list of tuples with variable and importance\n",
    "    feature_importances = [(feature, round(importance, 2)) for feature, importance in zip(feature_list, importances)]\n",
    "    feature_importances = sorted(feature_importances, key = lambda x: x[1], reverse = True)\n",
    "    [print('Variable: {:20} Importance: {}'.format(*pair)) for pair in feature_importances];\n",
    "    \n",
    "    return (test_features, test_labels, predictions, regressor)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label distribution: {0: 228606, 1: 10627}\n",
      "Training Features Shape: (179424, 8)\n",
      "Training Labels Shape: (179424,)\n",
      "Training label distribution: {0: 171430, 1: 7994}\n",
      "Testing Features Shape: (59809, 8)\n",
      "Testing Labels Shape: (59809,)\n",
      "Test label distribution: {0: 57176, 1: 2633}\n",
      "\n",
      "Predicted label distribution: {-0.5: 2, -0.4: 1, -0.3: 8, -0.2: 67, -0.1: 1800, 0.0: 39411, 0.1: 13802, 0.2: 2115, 0.3: 1178, 0.4: 681, 0.5: 429, 0.6: 315}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "56976 322\n",
      "0->1 1->0\n",
      "200 2311\n",
      "Accuracy: 0.96\n",
      "\n",
      "Variable: s2v_score            Importance: 0.44\n",
      "Variable: mass_match           Importance: 0.21\n",
      "Variable: cosine_score         Importance: 0.13\n",
      "Variable: cosine_matches       Importance: 0.0\n",
      "Variable: perc_mass_change     Importance: -0.0\n",
      "Variable: parent_mass          Importance: 0.0\n",
      "Variable: mod_cosine_score     Importance: -0.01\n",
      "Variable: mod_cosine_matches   Importance: -0.01\n"
     ]
    }
   ],
   "source": [
    "regression_match_perc_pmass = do_linear_regression(all_matches_s2v_match_perc_pmass, 'label')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
      "1556       0.000000               0          0.000000                   0   \n",
      "1782       0.000000               0          0.000008                   1   \n",
      "1828       0.000000               0          0.000000                   0   \n",
      "1875       0.000000               0          0.000000                   0   \n",
      "1901       0.000000               0          0.000000                   0   \n",
      "...             ...             ...               ...                 ...   \n",
      "89983      0.000000               0          0.000019                   1   \n",
      "89984      0.000000               0          0.000000                   0   \n",
      "91479      0.000007               1          0.000701                   3   \n",
      "91480      0.000012               1          0.000112                   2   \n",
      "91481      0.000021               1          0.000021                   1   \n",
      "\n",
      "       mass_match  s2v_score  label  \n",
      "1556          1.0  -0.088960      0  \n",
      "1782          1.0   0.104824      0  \n",
      "1828          1.0   0.030006      0  \n",
      "1875          1.0   0.120981      0  \n",
      "1901          1.0   0.046791      0  \n",
      "...           ...        ...    ...  \n",
      "89983         1.0   0.099174      0  \n",
      "89984         1.0   0.040209      0  \n",
      "91479         1.0   0.007186      0  \n",
      "91480         1.0   0.028249      0  \n",
      "91481         1.0   0.072551      0  \n",
      "\n",
      "[239233 rows x 7 columns]\n",
      "Label distribution: {0: 228606, 1: 10627}\n",
      "Training Features Shape: (179424, 6)\n",
      "Training Labels Shape: (179424,)\n",
      "Training label distribution: {0: 171430, 1: 7994}\n",
      "Testing Features Shape: (59809, 6)\n",
      "Testing Labels Shape: (59809,)\n",
      "Test label distribution: {0: 57176, 1: 2633}\n",
      "\n",
      "Predicted label distribution: {-0.4: 1, -0.3: 10, -0.2: 62, -0.1: 1103, -0.0: 42403, 0.1: 11521, 0.2: 2127, 0.3: 1189, 0.4: 682, 0.5: 679, 0.6: 32}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "56997 278\n",
      "0->1 1->0\n",
      "179 2355\n",
      "Accuracy: 0.96\n",
      "\n",
      "Variable: s2v_score            Importance: 0.41\n",
      "Variable: mass_match           Importance: 0.21\n",
      "Variable: cosine_score         Importance: 0.17\n",
      "Variable: cosine_matches       Importance: 0.01\n",
      "Variable: mod_cosine_matches   Importance: -0.01\n",
      "Variable: mod_cosine_score     Importance: -0.02\n"
     ]
    }
   ],
   "source": [
    "print(all_matches_s2v_labels)\n",
    "regression_labels = do_linear_regression(all_matches_s2v_labels, 'label')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label distribution: {0: 16122, 1: 3878}\n",
      "Training Features Shape: (15000, 6)\n",
      "Training Labels Shape: (15000,)\n",
      "Training label distribution: {0: 12115, 1: 2885}\n",
      "Testing Features Shape: (5000, 6)\n",
      "Testing Labels Shape: (5000,)\n",
      "Test label distribution: {0: 4007, 1: 993}\n",
      "\n",
      "Predicted label distribution: {-0.1: 147, 0.0: 1577, 0.1: 1701, 0.2: 293, 0.4: 14, 0.5: 294, 0.6: 778, 0.7: 196}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "3540 736\n",
      "0->1 1->0\n",
      "467 257\n",
      "Accuracy: 0.86\n",
      "\n",
      "Variable: mass_match           Importance: 0.47\n",
      "Variable: s2v_score            Importance: 0.29\n",
      "Variable: cosine_score         Importance: 0.06\n",
      "Variable: cosine_matches       Importance: 0.0\n",
      "Variable: mod_cosine_matches   Importance: -0.0\n",
      "Variable: mod_cosine_score     Importance: -0.03\n"
     ]
    }
   ],
   "source": [
    "regression_labels_top20 = do_linear_regression(top20_found_matches_s2v_labels, 'label')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label distribution: {0: 16122, 1: 3878}\n",
      "Training Features Shape: (15000, 8)\n",
      "Training Labels Shape: (15000,)\n",
      "Training label distribution: {0: 12115, 1: 2885}\n",
      "Testing Features Shape: (5000, 8)\n",
      "Testing Labels Shape: (5000,)\n",
      "Test label distribution: {0: 4007, 1: 993}\n",
      "\n",
      "Predicted label distribution: {-0.3: 1, -0.1: 141, -0.0: 1634, 0.1: 1599, 0.2: 343, 0.4: 7, 0.5: 299, 0.6: 789, 0.7: 187}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "3533 741\n",
      "0->1 1->0\n",
      "474 252\n",
      "Accuracy: 0.85\n",
      "\n",
      "Variable: mass_match           Importance: 0.48\n",
      "Variable: s2v_score            Importance: 0.25\n",
      "Variable: cosine_score         Importance: 0.07\n",
      "Variable: cosine_matches       Importance: 0.0\n",
      "Variable: mod_cosine_matches   Importance: -0.0\n",
      "Variable: perc_mass_change     Importance: -0.0\n",
      "Variable: parent_mass          Importance: 0.0\n",
      "Variable: mod_cosine_score     Importance: -0.02\n"
     ]
    }
   ],
   "source": [
    "found_matches_s2v_match_perc_pmass = find_info_matches(found_matches_s2v, documents_library_s2v, documents_query_s2v, calc_change=True, add_cols = ['parent_mass'], add_tanimoto_sim=False)\n",
    "top20_found_matches_s2v_match_perc_pmass = make_s2v_selection(found_matches_s2v_match_perc_pmass, 20)\n",
    "regression_labels_match_perc_pmass_top20 = do_linear_regression(top20_found_matches_s2v_match_perc_pmass, 'label')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label distribution: {0: 228606, 1: 228606}\n",
      "Training Features Shape: (342909, 6)\n",
      "Training Labels Shape: (342909,)\n",
      "Training label distribution: {0: 171294, 1: 171615}\n",
      "Testing Features Shape: (114303, 6)\n",
      "Testing Labels Shape: (114303,)\n",
      "Test label distribution: {0: 57312, 1: 56991}\n",
      "\n",
      "Predicted label distribution: {-2.0: 1, -1.9: 1, -1.8: 1, -1.7: 3, -1.6: 2, -1.5: 3, -1.4: 15, -1.3: 2, -1.2: 7, -1.1: 8, -1.0: 14, -0.9: 27, -0.8: 10, -0.7: 23, -0.6: 24, -0.5: 38, -0.4: 63, -0.3: 91, -0.2: 170, -0.1: 407, -0.0: 771, 0.1: 1825, 0.2: 10479, 0.3: 32479, 0.4: 21263, 0.5: 8827, 0.6: 6284, 0.7: 5614, 0.8: 6389, 0.9: 5973, 1.0: 6456, 1.1: 6908, 1.2: 102, 1.3: 23}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "51347 35653\n",
      "0->1 1->0\n",
      "5965 21338\n",
      "Accuracy: 0.76\n",
      "\n",
      "Variable: s2v_score            Importance: 0.98\n",
      "Variable: cosine_score         Importance: 0.26\n",
      "Variable: mass_match           Importance: 0.16\n",
      "Variable: cosine_matches       Importance: 0.04\n",
      "Variable: mod_cosine_matches   Importance: -0.04\n",
      "Variable: mod_cosine_score     Importance: -0.38\n"
     ]
    }
   ],
   "source": [
    "#regression on oversampled dataset\n",
    "regression_labels_oversampled = do_linear_regression(all_matches_s2v_oversampled, 'label')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Plot results querying top20"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {},
   "outputs": [],
   "source": [
    "#try selecting top 20 first\n",
    "\n",
    "test_m_s2v_top20 = []\n",
    "test_m_rf_all_top20 = []\n",
    "test_m_rf_top20_top20= []\n",
    "test_m_rf_top20_5050_top20= []\n",
    "test_m_rf_all_more_info_top20 = []\n",
    "test_m_rf_all_mm5_pmass_top20 = []\n",
    "test_m_rfreg_all_match_perc_top20 = []\n",
    "test_m_rfreg_all_match_perc_pmass_top20 = []\n",
    "test_m_reg_all_top20 = []\n",
    "test_m_reg_match_perc_pmass_top20 = []\n",
    "test_m_reg_top20_top20 = []\n",
    "test_m_rf_oversampled_top20 = []\n",
    "test_m_reg_oversampled_top20 = []\n",
    "\n",
    "threshold = 0.4\n",
    "topn = 20 #only look at top 20\n",
    "\n",
    "for ID in range(len(documents_query_s2v)):\n",
    "    inchikey_query = documents_query_s2v[ID]._obj.get(\"inchikey\")[:14]\n",
    "    #select top20\n",
    "    current_match = found_matches_s2v[ID].sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "\n",
    "    # Scenario 2: mass + sort by Spec2Vec\n",
    "    df_select = current_match[(current_match['mass_match'] == 1)\n",
    "                                  & (current_match['s2v_score'] > threshold)]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).index[0]\n",
    "        inchikey_match = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_match == inchikey_query)\n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_s2v_top20.append(best_bet)\n",
    "\n",
    "    #RF all data\n",
    "    predictions = all_result[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_rf_all_top20.append(best_bet)\n",
    "    \n",
    "    #RF with top20\n",
    "    predictions = results_top20[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_rf_top20_top20.append(best_bet)\n",
    "    \n",
    "    #RF with top20 50/50 split\n",
    "    predictions = results_top20_5050[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_rf_top20_5050_top20.append(best_bet)\n",
    "    \n",
    "    #RF all data with more mass match features\n",
    "    current_match_more_info = found_matches_s2v_more_info[ID].drop('label', axis = 1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rf_all_more_info[3].predict(current_match_more_info)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_more_info.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_rf_all_more_info_top20.append(best_bet)\n",
    "    \n",
    "    #RF all data\n",
    "    current_match_mm5_pmass = found_matches_s2v_mm5_pmass_sim[ID].drop(['similarity','label'],axis=1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = all_result_mm5_pmass[3].predict(current_match_mm5_pmass)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_mm5_pmass.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_rf_all_mm5_pmass_top20.append(best_bet)\n",
    "    \n",
    "    #RF regression all data with perc mass change\n",
    "    current_match_match_perc = found_matches_s2v_match_perc[ID].drop('label',axis=1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rfreg_all_match_perc[3].predict(current_match_match_perc)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_match_perc.iloc[predictions>=0.5]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_rfreg_all_match_perc_top20.append(best_bet)\n",
    "    \n",
    "    #RF regression all data with perc mass change and parent mass\n",
    "    current_match_match_perc_pmass = found_matches_s2v_match_perc_pmass_sim[ID].drop(['similarity','label'],axis=1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rfreg_all_match_perc_pmass[3].predict(current_match_match_perc_pmass)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_match_perc_pmass.iloc[predictions>=0.5]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_rfreg_all_match_perc_pmass_top20.append(best_bet)\n",
    "    \n",
    "    #regression all data\n",
    "    predictions = regression_labels[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    maxpred = predictions.argmax()\n",
    "    df_select = current_match.iloc[maxpred]\n",
    "    if predictions[maxpred] >= 0.5:\n",
    "        #best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        best_match_ID = df_select.name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_reg_all_top20.append(best_bet)\n",
    "    \n",
    "    #regression all data + mass match perc + parent mass\n",
    "    predictions = regression_match_perc_pmass[3].predict(current_match_match_perc_pmass)\n",
    "    #take the predicted match with highest s2v score\n",
    "    maxpred = predictions.argmax()\n",
    "    df_select = current_match_match_perc_pmass.iloc[maxpred]\n",
    "    if predictions[maxpred] >= 0.5:\n",
    "        #best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        best_match_ID = df_select.name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_reg_match_perc_pmass_top20.append(best_bet)\n",
    "    \n",
    "    #regression trained on top 20\n",
    "    predictions = regression_labels_top20[3].predict(current_match)\n",
    "    #take the predicted match with highest score above 0.5\n",
    "    maxpred = predictions.argmax()\n",
    "    df_select = current_match.iloc[maxpred]\n",
    "    if predictions[maxpred] >= 0.5:\n",
    "        #best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        best_match_ID = df_select.name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_reg_top20_top20.append(best_bet)\n",
    "    \n",
    "    #RF on oversampled data no added features\n",
    "    predictions = results_all_oversampled[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_rf_oversampled_top20.append(best_bet)\n",
    "    \n",
    "    #regression on oversampled data no added features\n",
    "    predictions = regression_labels_oversampled[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    maxpred = predictions.argmax()\n",
    "    df_select = current_match.iloc[maxpred]\n",
    "    if predictions[maxpred] >= 0.5:\n",
    "        #best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        best_match_ID = df_select.name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_m_reg_oversampled_top20.append(best_bet)\n",
    "    \n",
    "\n",
    "# Make arrays from lists:\n",
    "test_m_s2v_arr_top20 = np.array(test_m_s2v_top20)\n",
    "test_m_rf_all_arr_top20 = np.array(test_m_rf_all_top20)\n",
    "test_m_rf_top20_arr_top20 = np.array(test_m_rf_top20_top20)\n",
    "test_m_rf_top20_5050_arr_top20 = np.array(test_m_rf_top20_5050_top20)\n",
    "test_m_rf_all_more_info_arr_top20 = np.array(test_m_rf_all_more_info_top20)\n",
    "test_m_rf_all_mm5_pmass_arr_top20 = np.array(test_m_rf_all_mm5_pmass_top20)\n",
    "test_m_rfreg_all_match_perc_arr_top20 = np.array(test_m_rfreg_all_match_perc_top20)\n",
    "test_m_rfreg_all_match_perc_pmass_arr_top20 = np.array(test_m_rfreg_all_match_perc_pmass_top20)\n",
    "test_m_reg_all_arr_top20 = np.array(test_m_reg_all_top20)\n",
    "test_m_reg_match_perc_pmass_arr_top20 = np.array(test_m_reg_match_perc_pmass_top20)\n",
    "test_m_reg_top20_arr_top20 = np.array(test_m_reg_top20_top20)\n",
    "test_m_rf_oversampled_arr_top20 = np.array(test_m_rf_oversampled_top20)\n",
    "test_m_reg_oversampled_arr_top20 = np.array(test_m_reg_oversampled_top20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Spectral library matching comparison (>1 matching inchikey in library)')"
      ]
     },
     "execution_count": 149,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.style.use('ggplot')\n",
    "\n",
    "ylabels = ['mass + Spec2Vec',\n",
    "           'rf all',\n",
    "           'rf top20',\n",
    "           'rf top20 50/50-split',\n",
    "           'rf all + mass match 5',\n",
    "           'rf all + mass match 5 + pmass',\n",
    "           'rf regression + mass change perc',\n",
    "           'rf regression + mass change perc + pmass',\n",
    "           'regression all',\n",
    "           'regression + mass change perc + pmass',\n",
    "           'regression top 20',\n",
    "           'rf all oversampled',\n",
    "           'regression all oversampled']\n",
    "xlabels = ['true matches', 'false matches', 'no matches found']\n",
    "          #'mass + mod.cosine \\n + as backup: Spec2Vec + mod.cosine']\n",
    "data = np.array([[np.sum(test_m_s2v_arr_top20 == 1), np.sum(test_m_s2v_arr_top20 == 0), np.sum(test_m_s2v_arr_top20 == -1)],\n",
    "                [np.sum(test_m_rf_all_arr_top20 == 1), np.sum(test_m_rf_all_arr_top20 == 0), np.sum(test_m_rf_all_arr_top20 == -1)],\n",
    "                [np.sum(test_m_rf_top20_arr_top20 == 1), np.sum(test_m_rf_top20_arr_top20 == 0), np.sum(test_m_rf_top20_arr_top20 == -1)],\n",
    "                [np.sum(test_m_rf_top20_5050_arr_top20 == 1), np.sum(test_m_rf_top20_5050_arr_top20 == 0),\n",
    "                np.sum(test_m_rf_top20_5050_arr_top20 == -1)],\n",
    "                [np.sum(test_m_rf_all_more_info_arr_top20 == 1), np.sum(test_m_rf_all_more_info_arr_top20 == 0),\n",
    "                 np.sum(test_m_rf_all_more_info_arr_top20 == -1)],\n",
    "                [np.sum(test_m_rf_all_mm5_pmass_arr_top20 == 1), np.sum(test_m_rf_all_mm5_pmass_arr_top20 == 0),\n",
    "                 np.sum(test_m_rf_all_mm5_pmass_arr_top20 == -1)],\n",
    "                [np.sum(test_m_rfreg_all_match_perc_arr_top20 == 1), np.sum(test_m_rfreg_all_match_perc_arr_top20 == 0),\n",
    "                 np.sum(test_m_rfreg_all_match_perc_arr_top20 == -1)],\n",
    "                [np.sum(test_m_rfreg_all_match_perc_pmass_arr_top20 == 1), np.sum(test_m_rfreg_all_match_perc_pmass_arr_top20 == 0),\n",
    "                 np.sum(test_m_rfreg_all_match_perc_pmass_arr_top20 == -1)],\n",
    "                [np.sum(test_m_reg_all_arr_top20 == 1), np.sum(test_m_reg_all_arr_top20 == 0),\n",
    "                 np.sum(test_m_reg_all_arr_top20 == -1)],\n",
    "                [np.sum(test_m_reg_match_perc_pmass_arr_top20 == 1), np.sum(test_m_reg_match_perc_pmass_arr_top20 == 0),\n",
    "                 np.sum(test_m_reg_match_perc_pmass_arr_top20 == -1)],\n",
    "                [np.sum(test_m_reg_top20_arr_top20 == 1), np.sum(test_m_reg_top20_arr_top20 == 0),\n",
    "                 np.sum(test_m_reg_top20_arr_top20 == -1)],\n",
    "                [np.sum(test_m_rf_oversampled_arr_top20 == 1), np.sum(test_m_rf_oversampled_arr_top20 == 0),\n",
    "                 np.sum(test_m_rf_oversampled_arr_top20 == -1)],\n",
    "                [np.sum(test_m_reg_oversampled_arr_top20 == 1), np.sum(test_m_reg_oversampled_arr_top20 == 0),\n",
    "                 np.sum(test_m_reg_oversampled_arr_top20 == -1)]])\n",
    "\n",
    "\n",
    "dataframe = pd.DataFrame(data, columns=xlabels, index=ylabels)\n",
    "\n",
    "ax = dataframe.sort_values(by=['false matches'], ascending=True).plot.barh(stacked=True, edgecolor='none')\n",
    "\n",
    "horiz_offset = 1.03\n",
    "vert_offset = 1.\n",
    "ax.legend(bbox_to_anchor=(horiz_offset, vert_offset))\n",
    "plt.xlabel('Number of spectra')\n",
    "plt.title('Spectral library matching comparison (>1 matching inchikey in library)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Plot against new test set (new 1000 queries, queried on new model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 202,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C:\\Users\\joris\\Documents\\eScience_data\\data\\new_found_matches_s2v_match_perc_pmass_sim.pickle\n"
     ]
    }
   ],
   "source": [
    "#making new queries data in all different feature datasets\n",
    "outfile = os.path.join(path_data, 'new_found_matches_s2v_match_perc_pmass_sim.pickle')\n",
    "print(outfile)\n",
    "if os.path.exists(outfile):\n",
    "    with open(outfile, 'rb') as inf:\n",
    "        new_found_matches_s2v_match_perc_pmass_sim = pickle.load(inf)\n",
    "else:\n",
    "    new_found_matches_s2v_match_perc_pmass_sim = find_info_matches(new_found_matches_s2v, new_documents_library_s2v,\n",
    "                                                                   new_documents_query_s2v, calc_change=True,\n",
    "                                                                   add_cols = ['parent_mass'])\n",
    "    with open(outfile, 'wb') as outf:\n",
    "        pickle.dump(new_found_matches_s2v_match_perc_pmass_sim, outf)\n",
    "\n",
    "\n",
    "new_found_matches_s2v_mm5_pmass_sim = find_info_matches(new_found_matches_s2v, new_documents_library_s2v, new_documents_query_s2v, bins = [2, 5], add_cols = ['parent_mass'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>label</th>\n",
       "      <th>mass_match_2</th>\n",
       "      <th>mass_match_5</th>\n",
       "      <th>parent_mass</th>\n",
       "      <th>similarity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>46341</th>\n",
       "      <td>0.792853</td>\n",
       "      <td>3</td>\n",
       "      <td>0.792853</td>\n",
       "      <td>3</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.838697</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52743</th>\n",
       "      <td>0.766401</td>\n",
       "      <td>1</td>\n",
       "      <td>0.766401</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.589020</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>901.728724</td>\n",
       "      <td>0.901186</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34539</th>\n",
       "      <td>0.241939</td>\n",
       "      <td>1</td>\n",
       "      <td>0.955195</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.585604</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>833.760724</td>\n",
       "      <td>0.859649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47562</th>\n",
       "      <td>0.010289</td>\n",
       "      <td>1</td>\n",
       "      <td>0.010289</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.584888</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35469</th>\n",
       "      <td>0.243206</td>\n",
       "      <td>1</td>\n",
       "      <td>0.243206</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.565561</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>833.760724</td>\n",
       "      <td>0.859649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33568</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.560947</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>827.713724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52554</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.551094</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47158</th>\n",
       "      <td>0.285278</td>\n",
       "      <td>2</td>\n",
       "      <td>0.285278</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.550833</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>853.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47558</th>\n",
       "      <td>0.032219</td>\n",
       "      <td>1</td>\n",
       "      <td>0.032219</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.548738</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38527</th>\n",
       "      <td>0.013670</td>\n",
       "      <td>1</td>\n",
       "      <td>0.013693</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.545742</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>827.713724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52973</th>\n",
       "      <td>0.000008</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000008</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.544748</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38523</th>\n",
       "      <td>0.036623</td>\n",
       "      <td>1</td>\n",
       "      <td>0.040348</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.540438</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>801.697724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45757</th>\n",
       "      <td>0.766454</td>\n",
       "      <td>2</td>\n",
       "      <td>0.766454</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.538008</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>901.728724</td>\n",
       "      <td>0.901186</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>51823</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.537810</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43816</th>\n",
       "      <td>0.010825</td>\n",
       "      <td>1</td>\n",
       "      <td>0.384554</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.536827</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>859.775724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37029</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.534970</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>913.822724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37404</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.531860</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38017</th>\n",
       "      <td>0.000935</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000935</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.531845</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>857.760724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32638</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.531173</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>799.681724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37798</th>\n",
       "      <td>0.029850</td>\n",
       "      <td>1</td>\n",
       "      <td>0.033895</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.526315</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>885.791724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43970</th>\n",
       "      <td>0.634632</td>\n",
       "      <td>1</td>\n",
       "      <td>0.634632</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.521714</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47182</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.496593</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31878</th>\n",
       "      <td>0.000735</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000735</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.492990</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36538</th>\n",
       "      <td>0.000176</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000176</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.491255</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36319</th>\n",
       "      <td>0.000176</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000176</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.491255</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39997</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.486546</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39841</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.486546</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39051</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.478209</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39237</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.478209</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>51658</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.466711</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "46341      0.792853               3          0.792853                   3   \n",
       "52743      0.766401               1          0.766401                   1   \n",
       "34539      0.241939               1          0.955195                   2   \n",
       "47562      0.010289               1          0.010289                   1   \n",
       "35469      0.243206               1          0.243206                   1   \n",
       "33568      0.000000               0          0.000000                   0   \n",
       "52554      0.000000               0          0.000000                   0   \n",
       "47158      0.285278               2          0.285278                   2   \n",
       "47558      0.032219               1          0.032219                   1   \n",
       "38527      0.013670               1          0.013693                   2   \n",
       "52973      0.000008               1          0.000008                   1   \n",
       "38523      0.036623               1          0.040348                   2   \n",
       "45757      0.766454               2          0.766454                   2   \n",
       "51823      0.000000               0          0.000000                   0   \n",
       "43816      0.010825               1          0.384554                   3   \n",
       "37029      0.000000               0          0.000000                   0   \n",
       "37404      0.000000               0          0.000000                   0   \n",
       "38017      0.000935               1          0.000935                   1   \n",
       "32638      0.000000               0          0.000000                   0   \n",
       "37798      0.029850               1          0.033895                   3   \n",
       "43970      0.634632               1          0.634632                   1   \n",
       "47182      0.000000               0          0.000000                   0   \n",
       "31878      0.000735               1          0.000735                   1   \n",
       "36538      0.000176               1          0.000176                   1   \n",
       "36319      0.000176               1          0.000176                   1   \n",
       "39997      0.000000               0          0.000000                   0   \n",
       "39841      0.000000               0          0.000000                   0   \n",
       "39051      0.000000               0          0.000000                   0   \n",
       "39237      0.000000               0          0.000000                   0   \n",
       "51658      0.000000               0          0.000000                   0   \n",
       "\n",
       "       mass_match  s2v_score  label  mass_match_2  mass_match_5  parent_mass  \\\n",
       "46341         1.0   0.838697      0             1             1   829.728724   \n",
       "52743         0.0   0.589020      0             0             0   901.728724   \n",
       "34539         0.0   0.585604      0             0             1   833.760724   \n",
       "47562         1.0   0.584888      1             1             1   829.728724   \n",
       "35469         0.0   0.565561      0             0             1   833.760724   \n",
       "33568         0.0   0.560947      0             0             1   827.713724   \n",
       "52554         1.0   0.551094      0             1             1   829.728724   \n",
       "47158         0.0   0.550833      0             0             0   853.728724   \n",
       "47558         1.0   0.548738      0             1             1   829.728724   \n",
       "38527         0.0   0.545742      0             0             1   827.713724   \n",
       "52973         1.0   0.544748      1             1             1   829.728724   \n",
       "38523         0.0   0.540438      0             0             0   801.697724   \n",
       "45757         0.0   0.538008      0             0             0   901.728724   \n",
       "51823         1.0   0.537810      1             1             1   829.728724   \n",
       "43816         0.0   0.536827      0             0             0   859.775724   \n",
       "37029         0.0   0.534970      0             0             0   913.822724   \n",
       "37404         1.0   0.531860      0             1             1   829.728724   \n",
       "38017         0.0   0.531845      0             0             0   857.760724   \n",
       "32638         0.0   0.531173      0             0             0   799.681724   \n",
       "37798         0.0   0.526315      0             0             0   885.791724   \n",
       "43970         1.0   0.521714      1             1             1   829.728724   \n",
       "47182         1.0   0.496593      1             1             1   829.728724   \n",
       "31878         1.0   0.492990      0             1             1   829.728724   \n",
       "36538         1.0   0.491255      1             1             1   829.728724   \n",
       "36319         1.0   0.491255      1             1             1   829.728724   \n",
       "39997         1.0   0.486546      0             1             1   829.728724   \n",
       "39841         1.0   0.486546      0             1             1   829.728724   \n",
       "39051         1.0   0.478209      1             1             1   829.728724   \n",
       "39237         1.0   0.478209      1             1             1   829.728724   \n",
       "51658         1.0   0.466711      0             1             1   829.728724   \n",
       "\n",
       "       similarity  \n",
       "46341    0.947368  \n",
       "52743    0.901186  \n",
       "34539    0.859649  \n",
       "47562    1.000000  \n",
       "35469    0.859649  \n",
       "33568    1.000000  \n",
       "52554    0.947368  \n",
       "47158    1.000000  \n",
       "47558    0.947368  \n",
       "38527    1.000000  \n",
       "52973    1.000000  \n",
       "38523    0.947368  \n",
       "45757    0.901186  \n",
       "51823    1.000000  \n",
       "43816    0.947368  \n",
       "37029    1.000000  \n",
       "37404    0.947368  \n",
       "38017    0.947368  \n",
       "32638    1.000000  \n",
       "37798    0.947368  \n",
       "43970    1.000000  \n",
       "47182    1.000000  \n",
       "31878    0.947368  \n",
       "36538    1.000000  \n",
       "36319    1.000000  \n",
       "39997    0.947368  \n",
       "39841    0.947368  \n",
       "39051    1.000000  \n",
       "39237    1.000000  \n",
       "51658    0.947368  "
      ]
     },
     "execution_count": 155,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#dropping some columns from above to make all appropriate datasets\n",
    "new_found_matches_s2v_mm5_pmass_sim[0].sort_values(by=['s2v_score'], ascending=False).iloc[:30]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>label</th>\n",
       "      <th>mass_match_2</th>\n",
       "      <th>mass_match_5</th>\n",
       "      <th>parent_mass</th>\n",
       "      <th>similarity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>46341</th>\n",
       "      <td>0.792853</td>\n",
       "      <td>3</td>\n",
       "      <td>0.792853</td>\n",
       "      <td>3</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.838697</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52743</th>\n",
       "      <td>0.766401</td>\n",
       "      <td>1</td>\n",
       "      <td>0.766401</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.589020</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>901.728724</td>\n",
       "      <td>0.901186</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34539</th>\n",
       "      <td>0.241939</td>\n",
       "      <td>1</td>\n",
       "      <td>0.955195</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.585604</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>833.760724</td>\n",
       "      <td>0.859649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47562</th>\n",
       "      <td>0.010289</td>\n",
       "      <td>1</td>\n",
       "      <td>0.010289</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.584888</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35469</th>\n",
       "      <td>0.243206</td>\n",
       "      <td>1</td>\n",
       "      <td>0.243206</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.565561</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>833.760724</td>\n",
       "      <td>0.859649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33568</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.560947</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>827.713724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52554</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.551094</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47158</th>\n",
       "      <td>0.285278</td>\n",
       "      <td>2</td>\n",
       "      <td>0.285278</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.550833</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>853.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47558</th>\n",
       "      <td>0.032219</td>\n",
       "      <td>1</td>\n",
       "      <td>0.032219</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.548738</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38527</th>\n",
       "      <td>0.013670</td>\n",
       "      <td>1</td>\n",
       "      <td>0.013693</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.545742</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>827.713724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52973</th>\n",
       "      <td>0.000008</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000008</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.544748</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38523</th>\n",
       "      <td>0.036623</td>\n",
       "      <td>1</td>\n",
       "      <td>0.040348</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.540438</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>801.697724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45757</th>\n",
       "      <td>0.766454</td>\n",
       "      <td>2</td>\n",
       "      <td>0.766454</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.538008</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>901.728724</td>\n",
       "      <td>0.901186</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>51823</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.537810</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43816</th>\n",
       "      <td>0.010825</td>\n",
       "      <td>1</td>\n",
       "      <td>0.384554</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.536827</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>859.775724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37029</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.534970</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>913.822724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37404</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.531860</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38017</th>\n",
       "      <td>0.000935</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000935</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.531845</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>857.760724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32638</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.531173</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>799.681724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37798</th>\n",
       "      <td>0.029850</td>\n",
       "      <td>1</td>\n",
       "      <td>0.033895</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.526315</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>885.791724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43970</th>\n",
       "      <td>0.634632</td>\n",
       "      <td>1</td>\n",
       "      <td>0.634632</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.521714</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47182</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.496593</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31878</th>\n",
       "      <td>0.000735</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000735</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.492990</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36538</th>\n",
       "      <td>0.000176</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000176</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.491255</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36319</th>\n",
       "      <td>0.000176</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000176</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.491255</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39997</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.486546</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39841</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.486546</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39051</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.478209</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39237</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.478209</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>51658</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.466711</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>829.728724</td>\n",
       "      <td>0.947368</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "46341      0.792853               3          0.792853                   3   \n",
       "52743      0.766401               1          0.766401                   1   \n",
       "34539      0.241939               1          0.955195                   2   \n",
       "47562      0.010289               1          0.010289                   1   \n",
       "35469      0.243206               1          0.243206                   1   \n",
       "33568      0.000000               0          0.000000                   0   \n",
       "52554      0.000000               0          0.000000                   0   \n",
       "47158      0.285278               2          0.285278                   2   \n",
       "47558      0.032219               1          0.032219                   1   \n",
       "38527      0.013670               1          0.013693                   2   \n",
       "52973      0.000008               1          0.000008                   1   \n",
       "38523      0.036623               1          0.040348                   2   \n",
       "45757      0.766454               2          0.766454                   2   \n",
       "51823      0.000000               0          0.000000                   0   \n",
       "43816      0.010825               1          0.384554                   3   \n",
       "37029      0.000000               0          0.000000                   0   \n",
       "37404      0.000000               0          0.000000                   0   \n",
       "38017      0.000935               1          0.000935                   1   \n",
       "32638      0.000000               0          0.000000                   0   \n",
       "37798      0.029850               1          0.033895                   3   \n",
       "43970      0.634632               1          0.634632                   1   \n",
       "47182      0.000000               0          0.000000                   0   \n",
       "31878      0.000735               1          0.000735                   1   \n",
       "36538      0.000176               1          0.000176                   1   \n",
       "36319      0.000176               1          0.000176                   1   \n",
       "39997      0.000000               0          0.000000                   0   \n",
       "39841      0.000000               0          0.000000                   0   \n",
       "39051      0.000000               0          0.000000                   0   \n",
       "39237      0.000000               0          0.000000                   0   \n",
       "51658      0.000000               0          0.000000                   0   \n",
       "\n",
       "       mass_match  s2v_score  label  mass_match_2  mass_match_5  parent_mass  \\\n",
       "46341         1.0   0.838697      0             1             1   829.728724   \n",
       "52743         0.0   0.589020      0             0             0   901.728724   \n",
       "34539         0.0   0.585604      0             0             1   833.760724   \n",
       "47562         1.0   0.584888      1             1             1   829.728724   \n",
       "35469         0.0   0.565561      0             0             1   833.760724   \n",
       "33568         0.0   0.560947      0             0             1   827.713724   \n",
       "52554         1.0   0.551094      0             1             1   829.728724   \n",
       "47158         0.0   0.550833      0             0             0   853.728724   \n",
       "47558         1.0   0.548738      0             1             1   829.728724   \n",
       "38527         0.0   0.545742      0             0             1   827.713724   \n",
       "52973         1.0   0.544748      1             1             1   829.728724   \n",
       "38523         0.0   0.540438      0             0             0   801.697724   \n",
       "45757         0.0   0.538008      0             0             0   901.728724   \n",
       "51823         1.0   0.537810      1             1             1   829.728724   \n",
       "43816         0.0   0.536827      0             0             0   859.775724   \n",
       "37029         0.0   0.534970      0             0             0   913.822724   \n",
       "37404         1.0   0.531860      0             1             1   829.728724   \n",
       "38017         0.0   0.531845      0             0             0   857.760724   \n",
       "32638         0.0   0.531173      0             0             0   799.681724   \n",
       "37798         0.0   0.526315      0             0             0   885.791724   \n",
       "43970         1.0   0.521714      1             1             1   829.728724   \n",
       "47182         1.0   0.496593      1             1             1   829.728724   \n",
       "31878         1.0   0.492990      0             1             1   829.728724   \n",
       "36538         1.0   0.491255      1             1             1   829.728724   \n",
       "36319         1.0   0.491255      1             1             1   829.728724   \n",
       "39997         1.0   0.486546      0             1             1   829.728724   \n",
       "39841         1.0   0.486546      0             1             1   829.728724   \n",
       "39051         1.0   0.478209      1             1             1   829.728724   \n",
       "39237         1.0   0.478209      1             1             1   829.728724   \n",
       "51658         1.0   0.466711      0             1             1   829.728724   \n",
       "\n",
       "       similarity  \n",
       "46341    0.947368  \n",
       "52743    0.901186  \n",
       "34539    0.859649  \n",
       "47562    1.000000  \n",
       "35469    0.859649  \n",
       "33568    1.000000  \n",
       "52554    0.947368  \n",
       "47158    1.000000  \n",
       "47558    0.947368  \n",
       "38527    1.000000  \n",
       "52973    1.000000  \n",
       "38523    0.947368  \n",
       "45757    0.901186  \n",
       "51823    1.000000  \n",
       "43816    0.947368  \n",
       "37029    1.000000  \n",
       "37404    0.947368  \n",
       "38017    0.947368  \n",
       "32638    1.000000  \n",
       "37798    0.947368  \n",
       "43970    1.000000  \n",
       "47182    1.000000  \n",
       "31878    0.947368  \n",
       "36538    1.000000  \n",
       "36319    1.000000  \n",
       "39997    0.947368  \n",
       "39841    0.947368  \n",
       "39051    1.000000  \n",
       "39237    1.000000  \n",
       "51658    0.947368  "
      ]
     },
     "execution_count": 156,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "new_found_matches_s2v_mm5_pmass_sim[0].sort_values(by=['s2v_score'], ascending=False).iloc[:30]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "metadata": {},
   "outputs": [],
   "source": [
    "#try selecting top 40 first\n",
    "\n",
    "test_new_s2v_top40 = []\n",
    "test_new_rf_all_top40 = []\n",
    "test_new_rf_top20_top40= []\n",
    "test_new_rf_top20_5050_top40= []\n",
    "test_new_rf_all_more_info_top40 = []\n",
    "test_new_rf_all_mm5_pmass_top40 = []\n",
    "test_new_rfreg_all_match_perc_top40 = []\n",
    "test_new_rfreg_all_match_perc_pmass_top40 = []\n",
    "test_new_reg_all_top40 = []\n",
    "test_new_reg_match_perc_pmass_top40 = []\n",
    "test_new_reg_top20_top40 = []\n",
    "test_new_rf_oversampled_top40 = []\n",
    "test_new_reg_oversampled_top40 = []\n",
    "\n",
    "threshold = 0.4\n",
    "topn = 40 #only look at top 40\n",
    "\n",
    "for ID in range(len(new_documents_query_s2v)):\n",
    "    inchikey_query = new_documents_query_s2v[ID]._obj.get(\"inchikey\")[:14]\n",
    "    #select top20\n",
    "    current_match = new_found_matches_s2v[ID].sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "\n",
    "    # Scenario 2: mass + sort by Spec2Vec\n",
    "    df_select = current_match[(current_match['mass_match'] == 1)\n",
    "                                  & (current_match['s2v_score'] > threshold)]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).index[0]\n",
    "        inchikey_match = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_match == inchikey_query)\n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_s2v_top40.append(best_bet)\n",
    "\n",
    "    #RF all data\n",
    "    predictions = all_result[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rf_all_top40.append(best_bet)\n",
    "    \n",
    "    #RF with top20\n",
    "    predictions = results_top20[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rf_top20_top40.append(best_bet)\n",
    "    \n",
    "    #RF with top20 50/50 split\n",
    "    predictions = results_top20_5050[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rf_top20_5050_top40.append(best_bet)\n",
    "    \n",
    "    #RF all data with more mass match features 2 and 5\n",
    "    current_match_more_info = new_found_matches_s2v_mm5_pmass_sim[ID]\\\n",
    "        .drop(['label', 'parent_mass', 'similarity'], axis = 1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rf_all_more_info[3].predict(current_match_more_info)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_more_info.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rf_all_more_info_top40.append(best_bet)\n",
    "    \n",
    "    #RF all data mass match 5 and parent mass\n",
    "    current_match_mm5_pmass = new_found_matches_s2v_mm5_pmass_sim[ID]\\\n",
    "        .drop(['label', 'similarity', 'mass_match_2'], axis = 1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = all_result_mm5_pmass[3].predict(current_match_mm5_pmass)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_mm5_pmass.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rf_all_mm5_pmass_top40.append(best_bet)\n",
    "    \n",
    "    #RF regression all data with perc mass change\n",
    "    current_match_match_perc = new_found_matches_s2v_match_perc_pmass_sim[ID].drop(['similarity','label','parent_mass'],axis=1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rfreg_all_match_perc[3].predict(current_match_match_perc)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_match_perc.iloc[predictions>=0.5]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rfreg_all_match_perc_top40.append(best_bet)\n",
    "    \n",
    "    #RF regression all data with perc mass change and parent mass\n",
    "    current_match_match_perc_pmass = new_found_matches_s2v_match_perc_pmass_sim[ID].drop(['similarity','label'],axis=1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rfreg_all_match_perc_pmass[3].predict(current_match_match_perc_pmass)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_match_perc_pmass.iloc[predictions>=0.5]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rfreg_all_match_perc_pmass_top40.append(best_bet)\n",
    "    \n",
    "    #regression all data\n",
    "    predictions = regression_labels[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    maxpred = predictions.argmax()\n",
    "    df_select = current_match.iloc[maxpred]\n",
    "    if predictions[maxpred] >= 0.5:\n",
    "        #best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        best_match_ID = df_select.name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_reg_all_top40.append(best_bet)\n",
    "    \n",
    "    #regression all data + mass match perc + parent mass\n",
    "    predictions = regression_match_perc_pmass[3].predict(current_match_match_perc_pmass)\n",
    "    #take the predicted match with highest s2v score\n",
    "    maxpred = predictions.argmax()\n",
    "    df_select = current_match_match_perc_pmass.iloc[maxpred]\n",
    "    if predictions[maxpred] >= 0.5:\n",
    "        #best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        best_match_ID = df_select.name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_reg_match_perc_pmass_top40.append(best_bet)\n",
    "    \n",
    "    #regression trained on top 20\n",
    "    predictions = regression_labels_top20[3].predict(current_match)\n",
    "    #take the predicted match with highest score above 0.5\n",
    "    maxpred = predictions.argmax()\n",
    "    df_select = current_match.iloc[maxpred]\n",
    "    if predictions[maxpred] >= 0.5:\n",
    "        #best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        best_match_ID = df_select.name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_reg_top20_top40.append(best_bet)\n",
    "    \n",
    "    #RF on oversampled data no added features\n",
    "    predictions = results_all_oversampled[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rf_oversampled_top40.append(best_bet)\n",
    "    \n",
    "    #regression on oversampled data no added features\n",
    "    predictions = regression_labels_oversampled[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    maxpred = predictions.argmax()\n",
    "    df_select = current_match.iloc[maxpred]\n",
    "    if predictions[maxpred] >= 0.5:\n",
    "        #best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        best_match_ID = df_select.name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_reg_oversampled_top40.append(best_bet)\n",
    "\n",
    "# Make arrays from lists:\n",
    "test_new_s2v_arr_top40 = np.array(test_new_s2v_top40)\n",
    "test_new_rf_all_arr_top40 = np.array(test_new_rf_all_top40)\n",
    "test_new_rf_top20_arr_top40 = np.array(test_new_rf_top20_top40)\n",
    "test_new_rf_top20_5050_arr_top40 = np.array(test_new_rf_top20_5050_top40)\n",
    "test_new_rf_all_more_info_arr_top40 = np.array(test_new_rf_all_more_info_top40)\n",
    "test_new_rf_all_mm5_pmass_arr_top40 = np.array(test_new_rf_all_mm5_pmass_top40)\n",
    "test_new_rfreg_all_match_perc_arr_top40 = np.array(test_new_rfreg_all_match_perc_top40)\n",
    "test_new_rfreg_all_match_perc_pmass_arr_top40 = np.array(test_new_rfreg_all_match_perc_pmass_top40)\n",
    "test_new_reg_all_arr_top40 = np.array(test_new_reg_all_top40)\n",
    "test_new_reg_match_perc_pmass_arr_top40 = np.array(test_new_reg_match_perc_pmass_top40)\n",
    "test_new_reg_top20_arr_top40 = np.array(test_new_reg_top20_top40)\n",
    "test_new_rf_oversampled_arr_top40 = np.array(test_new_rf_oversampled_top40)\n",
    "test_new_reg_oversampled_arr_top40 = np.array(test_new_reg_oversampled_top40)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 158,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Spectral library matching comparison new test set looking at top 40\\n(>5 matching inchikey in library)')"
      ]
     },
     "execution_count": 158,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.style.use('ggplot')\n",
    "\n",
    "ylabels = ['mass + Spec2Vec',\n",
    "           'rf all',\n",
    "           'rf top20',\n",
    "           'rf top20 50/50-split',\n",
    "           'rf all + mass match 5',\n",
    "           'rf all + mass match 5 + pmass',\n",
    "           'rf regression + mass change perc',\n",
    "           'rf regression + mass change perc + pmass',\n",
    "           'regression all',\n",
    "           'regression + mass change perc + pmass',\n",
    "           'regression top 20',\n",
    "           'rf all oversampled',\n",
    "           'regression all oversampled']\n",
    "xlabels = ['true matches', 'false matches', 'no matches found']\n",
    "          #'mass + mod.cosine \\n + as backup: Spec2Vec + mod.cosine']\n",
    "data = np.array([[np.sum(test_new_s2v_arr_top40 == 1), np.sum(test_new_s2v_arr_top40 == 0), np.sum(test_new_s2v_arr_top40 == -1)],\n",
    "                [np.sum(test_new_rf_all_arr_top40 == 1), np.sum(test_new_rf_all_arr_top40 == 0), np.sum(test_new_rf_all_arr_top40 == -1)],\n",
    "                [np.sum(test_new_rf_top20_arr_top40 == 1), np.sum(test_new_rf_top20_arr_top40 == 0), np.sum(test_new_rf_top20_arr_top40 == -1)],\n",
    "                [np.sum(test_new_rf_top20_5050_arr_top40 == 1), np.sum(test_new_rf_top20_5050_arr_top40 == 0),\n",
    "                np.sum(test_new_rf_top20_5050_arr_top40 == -1)],\n",
    "                [np.sum(test_new_rf_all_more_info_arr_top40 == 1), np.sum(test_new_rf_all_more_info_arr_top40 == 0),\n",
    "                 np.sum(test_new_rf_all_more_info_arr_top40 == -1)],\n",
    "                [np.sum(test_new_rf_all_mm5_pmass_arr_top40 == 1), np.sum(test_new_rf_all_mm5_pmass_arr_top40 == 0),\n",
    "                 np.sum(test_new_rf_all_mm5_pmass_arr_top40 == -1)],\n",
    "                [np.sum(test_new_rfreg_all_match_perc_arr_top40 == 1), np.sum(test_new_rfreg_all_match_perc_arr_top40 == 0),\n",
    "                 np.sum(test_new_rfreg_all_match_perc_arr_top40 == -1)],\n",
    "                [np.sum(test_new_rfreg_all_match_perc_pmass_arr_top40 == 1), np.sum(test_new_rfreg_all_match_perc_pmass_arr_top40 == 0),\n",
    "                 np.sum(test_new_rfreg_all_match_perc_pmass_arr_top40 == -1)],\n",
    "                [np.sum(test_new_reg_all_arr_top40 == 1), np.sum(test_new_reg_all_arr_top40 == 0),\n",
    "                 np.sum(test_new_reg_all_arr_top40 == -1)],\n",
    "                [np.sum(test_new_reg_match_perc_pmass_arr_top40 == 1), np.sum(test_new_reg_match_perc_pmass_arr_top40 == 0),\n",
    "                 np.sum(test_new_reg_match_perc_pmass_arr_top40 == -1)],\n",
    "                [np.sum(test_new_reg_top20_arr_top40 == 1), np.sum(test_new_reg_top20_arr_top40 == 0),\n",
    "                 np.sum(test_new_reg_top20_arr_top40 == -1)],\n",
    "                [np.sum(test_new_rf_oversampled_arr_top40 == 1), np.sum(test_new_rf_oversampled_arr_top40 == 0),\n",
    "                 np.sum(test_new_rf_oversampled_arr_top40 == -1)],\n",
    "                [np.sum(test_new_reg_oversampled_arr_top40 == 1), np.sum(test_new_reg_oversampled_arr_top40 == 0),\n",
    "                 np.sum(test_new_reg_oversampled_arr_top40 == -1)]])\n",
    "\n",
    "\n",
    "\n",
    "dataframe = pd.DataFrame(data, columns=xlabels, index=ylabels)\n",
    "\n",
    "ax = dataframe.sort_values(by=['false matches'], ascending=True).plot.barh(stacked=True, edgecolor='none')\n",
    "\n",
    "horiz_offset = 1.03\n",
    "vert_offset = 1.\n",
    "ax.legend(bbox_to_anchor=(horiz_offset, vert_offset))\n",
    "plt.xlabel('Number of spectra')\n",
    "plt.title('Spectral library matching comparison new test set looking at top 40\\n(>5 matching inchikey in library)')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "metadata": {},
   "outputs": [],
   "source": [
    "#try selecting top 20 first\n",
    "\n",
    "test_new_s2v_top20 = []\n",
    "test_new_rf_all_top20 = []\n",
    "test_new_rf_top20_top20= []\n",
    "test_new_rf_top20_5050_top20= []\n",
    "test_new_rf_all_more_info_top20 = []\n",
    "test_new_rf_all_mm5_pmass_top20 = []\n",
    "test_new_rfreg_all_match_perc_top20 = []\n",
    "test_new_rfreg_all_match_perc_pmass_top20 = []\n",
    "test_new_reg_all_top20 = []\n",
    "test_new_reg_match_perc_pmass_top20 = []\n",
    "test_new_reg_top20_top20 = []\n",
    "test_new_rf_oversampled_top20 = []\n",
    "test_new_reg_oversampled_top20 = []\n",
    "\n",
    "threshold = 0.4\n",
    "topn = 20 #only look at top 20\n",
    "\n",
    "for ID in range(len(new_documents_query_s2v)):\n",
    "    inchikey_query = new_documents_query_s2v[ID]._obj.get(\"inchikey\")[:14]\n",
    "    #select top20\n",
    "    current_match = new_found_matches_s2v[ID].sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "\n",
    "    # Scenario 2: mass + sort by Spec2Vec\n",
    "    df_select = current_match[(current_match['mass_match'] == 1)\n",
    "                                  & (current_match['s2v_score'] > threshold)]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).index[0]\n",
    "        inchikey_match = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_match == inchikey_query)\n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_s2v_top20.append(best_bet)\n",
    "\n",
    "    #RF all data\n",
    "    predictions = all_result[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rf_all_top20.append(best_bet)\n",
    "    \n",
    "    #RF with top20\n",
    "    predictions = results_top20[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rf_top20_top20.append(best_bet)\n",
    "    \n",
    "    #RF with top20 50/50 split\n",
    "    predictions = results_top20_5050[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rf_top20_5050_top20.append(best_bet)\n",
    "    \n",
    "    #RF all data with more mass match features 2 and 5\n",
    "    current_match_more_info = new_found_matches_s2v_mm5_pmass_sim[ID]\\\n",
    "        .drop(['label', 'parent_mass', 'similarity'], axis = 1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rf_all_more_info[3].predict(current_match_more_info)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_more_info.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rf_all_more_info_top20.append(best_bet)\n",
    "    \n",
    "    #RF all data mass match 5 and parent mass\n",
    "    current_match_mm5_pmass = new_found_matches_s2v_mm5_pmass_sim[ID]\\\n",
    "        .drop(['label', 'similarity', 'mass_match_2'], axis = 1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = all_result_mm5_pmass[3].predict(current_match_mm5_pmass)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_mm5_pmass.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rf_all_mm5_pmass_top20.append(best_bet)\n",
    "    \n",
    "    #RF regression all data with perc mass change\n",
    "    current_match_match_perc = new_found_matches_s2v_match_perc_pmass_sim[ID].drop(['similarity','label','parent_mass'],axis=1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rfreg_all_match_perc[3].predict(current_match_match_perc)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_match_perc.iloc[predictions>=0.5]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rfreg_all_match_perc_top20.append(best_bet)\n",
    "    \n",
    "    #RF regression all data with perc mass change and parent mass\n",
    "    current_match_match_perc_pmass = new_found_matches_s2v_match_perc_pmass_sim[ID].drop(['similarity','label'],axis=1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rfreg_all_match_perc_pmass[3].predict(current_match_match_perc_pmass)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_match_perc_pmass.iloc[predictions>=0.5]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rfreg_all_match_perc_pmass_top20.append(best_bet)\n",
    "    \n",
    "    #regression all data\n",
    "    predictions = regression_labels[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    maxpred = predictions.argmax()\n",
    "    df_select = current_match.iloc[maxpred]\n",
    "    if predictions[maxpred] >= 0.5:\n",
    "        #best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        best_match_ID = df_select.name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_reg_all_top20.append(best_bet)\n",
    "    \n",
    "    #regression all data + mass match perc + parent mass\n",
    "    predictions = regression_match_perc_pmass[3].predict(current_match_match_perc_pmass)\n",
    "    #take the predicted match with highest s2v score\n",
    "    maxpred = predictions.argmax()\n",
    "    df_select = current_match_match_perc_pmass.iloc[maxpred]\n",
    "    if predictions[maxpred] >= 0.5:\n",
    "        #best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        best_match_ID = df_select.name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_reg_match_perc_pmass_top20.append(best_bet)\n",
    "    \n",
    "    #regression trained on top 20\n",
    "    predictions = regression_labels_top20[3].predict(current_match)\n",
    "    #take the predicted match with highest score above 0.5\n",
    "    maxpred = predictions.argmax()\n",
    "    df_select = current_match.iloc[maxpred]\n",
    "    if predictions[maxpred] >= 0.5:\n",
    "        #best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        best_match_ID = df_select.name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_reg_top20_top20.append(best_bet)\n",
    "    \n",
    "    #RF on oversampled data no added features\n",
    "    predictions = results_all_oversampled[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match.iloc[predictions==1]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rf_oversampled_top20.append(best_bet)\n",
    "    \n",
    "    #regression on oversampled data no added features\n",
    "    predictions = regression_labels_oversampled[3].predict(current_match)\n",
    "    #take the predicted match with highest s2v score\n",
    "    maxpred = predictions.argmax()\n",
    "    df_select = current_match.iloc[maxpred]\n",
    "    if predictions[maxpred] >= 0.5:\n",
    "        #best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        best_match_ID = df_select.name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_reg_oversampled_top20.append(best_bet)\n",
    "\n",
    "# Make arrays from lists:\n",
    "test_new_s2v_arr_top20 = np.array(test_new_s2v_top20)\n",
    "test_new_rf_all_arr_top20 = np.array(test_new_rf_all_top20)\n",
    "test_new_rf_top20_arr_top20 = np.array(test_new_rf_top20_top20)\n",
    "test_new_rf_top20_5050_arr_top20 = np.array(test_new_rf_top20_5050_top20)\n",
    "test_new_rf_all_more_info_arr_top20 = np.array(test_new_rf_all_more_info_top20)\n",
    "test_new_rf_all_mm5_pmass_arr_top20 = np.array(test_new_rf_all_mm5_pmass_top20)\n",
    "test_new_rfreg_all_match_perc_arr_top20 = np.array(test_new_rfreg_all_match_perc_top20)\n",
    "test_new_rfreg_all_match_perc_pmass_arr_top20 = np.array(test_new_rfreg_all_match_perc_pmass_top20)\n",
    "test_new_reg_all_arr_top20 = np.array(test_new_reg_all_top20)\n",
    "test_new_reg_match_perc_pmass_arr_top20 = np.array(test_new_reg_match_perc_pmass_top20)\n",
    "test_new_reg_top20_arr_top20 = np.array(test_new_reg_top20_top20)\n",
    "test_new_rf_oversampled_arr_top20 = np.array(test_new_rf_oversampled_top20)\n",
    "test_new_reg_oversampled_arr_top20 = np.array(test_new_reg_oversampled_top20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Spectral library matching comparison new test set looking at top 20\\n(>5 matching inchikey in library)')"
      ]
     },
     "execution_count": 160,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.style.use('ggplot')\n",
    "\n",
    "ylabels = ['mass + Spec2Vec',\n",
    "           'rf all',\n",
    "           'rf top20',\n",
    "           'rf top20 50/50-split',\n",
    "           'rf all + mass match 5',\n",
    "           'rf all + mass match 5 + pmass',\n",
    "           'rf regression + mass change perc',\n",
    "           'rf regression + mass change perc + pmass',\n",
    "           'regression all',\n",
    "           'regression + mass change perc + pmass',\n",
    "           'regression top 20',\n",
    "           'rf all oversampled',\n",
    "           'regression all oversampled']\n",
    "xlabels = ['true matches', 'false matches', 'no matches found']\n",
    "          #'mass + mod.cosine \\n + as backup: Spec2Vec + mod.cosine']\n",
    "data = np.array([[np.sum(test_new_s2v_arr_top20 == 1), np.sum(test_new_s2v_arr_top20 == 0), np.sum(test_new_s2v_arr_top20 == -1)],\n",
    "                [np.sum(test_new_rf_all_arr_top20 == 1), np.sum(test_new_rf_all_arr_top20 == 0), np.sum(test_new_rf_all_arr_top20 == -1)],\n",
    "                [np.sum(test_new_rf_top20_arr_top20 == 1), np.sum(test_new_rf_top20_arr_top20 == 0), np.sum(test_new_rf_top20_arr_top20 == -1)],\n",
    "                [np.sum(test_new_rf_top20_5050_arr_top20 == 1), np.sum(test_new_rf_top20_5050_arr_top20 == 0),\n",
    "                np.sum(test_new_rf_top20_5050_arr_top20 == -1)],\n",
    "                [np.sum(test_new_rf_all_more_info_arr_top20 == 1), np.sum(test_new_rf_all_more_info_arr_top20 == 0),\n",
    "                 np.sum(test_new_rf_all_more_info_arr_top20 == -1)],\n",
    "                [np.sum(test_new_rf_all_mm5_pmass_arr_top20 == 1), np.sum(test_new_rf_all_mm5_pmass_arr_top20 == 0),\n",
    "                 np.sum(test_new_rf_all_mm5_pmass_arr_top20 == -1)],\n",
    "                [np.sum(test_new_rfreg_all_match_perc_arr_top20 == 1), np.sum(test_new_rfreg_all_match_perc_arr_top20 == 0),\n",
    "                 np.sum(test_new_rfreg_all_match_perc_arr_top20 == -1)],\n",
    "                [np.sum(test_new_rfreg_all_match_perc_pmass_arr_top20 == 1), np.sum(test_new_rfreg_all_match_perc_pmass_arr_top20 == 0),\n",
    "                 np.sum(test_new_rfreg_all_match_perc_pmass_arr_top20 == -1)],\n",
    "                [np.sum(test_new_reg_all_arr_top20 == 1), np.sum(test_new_reg_all_arr_top20 == 0),\n",
    "                 np.sum(test_new_reg_all_arr_top20 == -1)],\n",
    "                [np.sum(test_new_reg_match_perc_pmass_arr_top20 == 1), np.sum(test_new_reg_match_perc_pmass_arr_top20 == 0),\n",
    "                 np.sum(test_new_reg_match_perc_pmass_arr_top20 == -1)],\n",
    "                [np.sum(test_new_reg_top20_arr_top20 == 1), np.sum(test_new_reg_top20_arr_top20 == 0),\n",
    "                 np.sum(test_new_reg_top20_arr_top20 == -1)],\n",
    "                [np.sum(test_new_rf_oversampled_arr_top20 == 1), np.sum(test_new_rf_oversampled_arr_top20 == 0),\n",
    "                 np.sum(test_new_rf_oversampled_arr_top20 == -1)],\n",
    "                [np.sum(test_new_reg_oversampled_arr_top20 == 1), np.sum(test_new_reg_oversampled_arr_top20 == 0),\n",
    "                 np.sum(test_new_reg_oversampled_arr_top20 == -1)]])\n",
    "\n",
    "\n",
    "\n",
    "dataframe = pd.DataFrame(data, columns=xlabels, index=ylabels)\n",
    "\n",
    "ax = dataframe.sort_values(by=['false matches'], ascending=True).plot.barh(stacked=True, edgecolor='none')\n",
    "\n",
    "horiz_offset = 1.03\n",
    "vert_offset = 1.\n",
    "ax.legend(bbox_to_anchor=(horiz_offset, vert_offset))\n",
    "plt.xlabel('Number of spectra')\n",
    "plt.title('Spectral library matching comparison new test set looking at top 20\\n(>5 matching inchikey in library)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Trying different test sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {},
   "outputs": [],
   "source": [
    "new2_found_matches_s2v_match_perc_pmass_sim = find_info_matches(new2_found_matches_s2v, new2_documents_library_s2v, new2_documents_query_s2v, calc_change=True, add_cols = ['parent_mass'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {},
   "outputs": [],
   "source": [
    "new3_found_matches_s2v_match_perc_pmass_sim = find_info_matches(new3_found_matches_s2v, new3_documents_library_s2v, new3_documents_query_s2v, calc_change=True, add_cols = ['parent_mass'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "metadata": {},
   "outputs": [],
   "source": [
    "old_and_unique_found_matches_s2v_match_perc_pmass_sim = find_info_matches(old_and_unique_found_matches_s2v, old_and_unique_documents_library_s2v, old_and_unique_documents_query_s2v, calc_change=True, add_cols = ['parent_mass'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label distribution: {0: 231903, 1: 9151}\n",
      "Training Features Shape: (180790, 8)\n",
      "Training Labels Shape: (180790,)\n",
      "Training label distribution: {0: 173916, 1: 6874}\n",
      "Testing Features Shape: (60264, 8)\n",
      "Testing Labels Shape: (60264,)\n",
      "Test label distribution: {0: 57987, 1: 2277}\n",
      "\n",
      "Predicted label distribution: {0.0: 55682, 0.1: 1207, 0.2: 530, 0.3: 399, 0.4: 286, 0.5: 274, 0.6: 285, 0.7: 272, 0.8: 278, 0.9: 381, 1.0: 670}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "57617 1640\n",
      "0->1 1->0\n",
      "370 637\n",
      "Accuracy: 0.98\n",
      "\n",
      "Variable: perc_mass_change     Importance: 0.44\n",
      "Variable: parent_mass          Importance: 0.2\n",
      "Variable: s2v_score            Importance: 0.16\n",
      "Variable: cosine_score         Importance: 0.07\n",
      "Variable: mod_cosine_score     Importance: 0.06\n",
      "Variable: cosine_matches       Importance: 0.04\n",
      "Variable: mod_cosine_matches   Importance: 0.04\n",
      "Variable: mass_match           Importance: 0.0\n"
     ]
    }
   ],
   "source": [
    "#make a new rfreg model from the first new training set\n",
    "all_new_matches_s2v_match_perc_pmass = new_found_matches_s2v_match_perc_pmass_sim[0].append(new_found_matches_s2v_match_perc_pmass_sim[1:]).drop('similarity',axis=1)\n",
    "rfreg_all_new_match_perc_pmass = do_random_forest(all_new_matches_s2v_match_perc_pmass, 'label', rf_regr = True,  verbose = True, num_est = 500)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "metadata": {},
   "outputs": [],
   "source": [
    "#This is the new rf reg model on old and new test data\n",
    "\n",
    "#look at top20\n",
    "test_old_rfreg_new_all_match_perc_pmass_top20 = []\n",
    "test_new_rfreg_new_all_match_perc_pmass_top20 = []\n",
    "threshold = 0.4\n",
    "topn = 20 #only look at top 20\n",
    "\n",
    "for ID in range(len(new_documents_query_s2v)): #I know they have the same length\n",
    "    inchikey_query = documents_query_s2v[ID]._obj.get(\"inchikey\")[:14]\n",
    "    #select top20\n",
    "    current_match = found_matches_s2v[ID].sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "\n",
    "    #RF regression all data with perc mass change and parent mass, new rfreg model, old test set\n",
    "    current_match_match_perc_pmass_old = found_matches_s2v_match_perc_pmass_sim[ID].drop(['similarity','label'],axis=1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rfreg_all_new_match_perc_pmass[3].predict(current_match_match_perc_pmass_old)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_match_perc_pmass_old.iloc[predictions>=0.5]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_old_rfreg_new_all_match_perc_pmass_top20.append(best_bet)\n",
    "    \n",
    "    #RF regression all data with perc mass change and parent mass, new rfreg model, new test set\n",
    "    inchikey_query = new_documents_query_s2v[ID]._obj.get(\"inchikey\")[:14]\n",
    "    #select top20\n",
    "    current_match = new_found_matches_s2v[ID].sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    \n",
    "    current_match_match_perc_pmass_new = new_found_matches_s2v_match_perc_pmass_sim[ID].drop(['similarity','label'],axis=1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rfreg_all_new_match_perc_pmass[3].predict(current_match_match_perc_pmass_new)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_match_perc_pmass_new.iloc[predictions>=0.5]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new_rfreg_new_all_match_perc_pmass_top20.append(best_bet)\n",
    "\n",
    "#make arrays from lists\n",
    "test_old_rfreg_new_all_match_perc_pmass_arr_top20 = np.array(test_old_rfreg_new_all_match_perc_pmass_top20)\n",
    "test_new_rfreg_new_all_match_perc_pmass_arr_top20 = np.array(test_new_rfreg_new_all_match_perc_pmass_top20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 210,
   "metadata": {},
   "outputs": [],
   "source": [
    "#look at top20\n",
    "test_new2_s2v_top20 = []\n",
    "test_new2_rfreg_all_match_perc_pmass_top20 = []\n",
    "test_new2_s2v_same_mass_top20 = []\n",
    "threshold = 0.4\n",
    "topn = 20 #only look at top 20\n",
    "\n",
    "for ID in range(len(new2_documents_query_s2v)):\n",
    "    inchikey_query = new2_documents_query_s2v[ID]._obj.get(\"inchikey\")[:14]\n",
    "    #select top20\n",
    "    current_match = new2_found_matches_s2v[ID].sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "\n",
    "    # Scenario 2: mass + sort by Spec2Vec\n",
    "    df_select = current_match[(current_match['mass_match'] == 1)\n",
    "                                  & (current_match['s2v_score'] > threshold)]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).index[0]\n",
    "        inchikey_match = new2_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_match == inchikey_query)\n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new2_s2v_top20.append(best_bet)\n",
    "\n",
    "    #RF regression all data with perc mass change and parent mass\n",
    "    current_match_match_perc_pmass = new2_found_matches_s2v_match_perc_pmass_sim[ID].drop(['similarity','label'],axis=1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rfreg_all_match_perc_pmass[3].predict(current_match_match_perc_pmass)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_match_perc_pmass.iloc[predictions>=0.5]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new2_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new2_rfreg_all_match_perc_pmass_top20.append(best_bet)\n",
    "    \n",
    "    # mass change 0 + sort by Spec2Vec\n",
    "    df_select = current_match_match_perc_pmass[(current_match_match_perc_pmass['perc_mass_change'] == 0)\n",
    "                                  & (current_match_match_perc_pmass['s2v_score'] > threshold)]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).index[0]\n",
    "        inchikey_match = new2_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_match == inchikey_query)\n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new2_s2v_same_mass_top20.append(best_bet)\n",
    "\n",
    "#make arrays from lists\n",
    "test_new2_s2v_arr_top20 = np.array(test_new2_s2v_top20)\n",
    "test_new2_rfreg_all_match_perc_pmass_arr_top20 = np.array(test_new2_rfreg_all_match_perc_pmass_top20)\n",
    "test_new2_s2v_same_mass_arr_top20 = np.array(test_new2_s2v_same_mass_top20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 207,
   "metadata": {},
   "outputs": [],
   "source": [
    "#look at top20\n",
    "test_new3_s2v_top20 = []\n",
    "test_new3_rfreg_all_match_perc_pmass_top20 = []\n",
    "threshold = 0.4\n",
    "topn = 20 #only look at top 20\n",
    "\n",
    "for ID in range(len(new3_documents_query_s2v)):\n",
    "    inchikey_query = new3_documents_query_s2v[ID]._obj.get(\"inchikey\")[:14]\n",
    "    #select top20\n",
    "    current_match = new3_found_matches_s2v[ID].sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "\n",
    "    # Scenario 2: mass + sort by Spec2Vec\n",
    "    df_select = current_match[(current_match['mass_match'] == 1)\n",
    "                                  & (current_match['s2v_score'] > threshold)]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).index[0]\n",
    "        inchikey_match = new3_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_match == inchikey_query)\n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new3_s2v_top20.append(best_bet)\n",
    "\n",
    "    #RF regression all data with perc mass change and parent mass\n",
    "    current_match_match_perc_pmass = new3_found_matches_s2v_match_perc_pmass_sim[ID].drop(['similarity','label'],axis=1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rfreg_all_match_perc_pmass[3].predict(current_match_match_perc_pmass)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_match_perc_pmass.iloc[predictions>=0.5]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = new3_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_new3_rfreg_all_match_perc_pmass_top20.append(best_bet)\n",
    "\n",
    "#make arrays from lists\n",
    "test_new3_s2v_arr_top20 = np.array(test_new3_s2v_top20)\n",
    "test_new3_rfreg_all_match_perc_pmass_arr_top20 = np.array(test_new3_rfreg_all_match_perc_pmass_top20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {},
   "outputs": [],
   "source": [
    "#what if you just take the top hit above a threshold\n",
    "#look at top20\n",
    "test_top_hit_s2v_top20 = []\n",
    "threshold = 0.4\n",
    "topn = 20 #only look at top 20\n",
    "\n",
    "for ID in range(len(documents_query_s2v)):\n",
    "    inchikey_query = documents_query_s2v[ID]._obj.get(\"inchikey\")[:14]\n",
    "    #select top20\n",
    "    current_match = found_matches_s2v[ID].sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "\n",
    "    # Scenario 2: Spec2Vec + threshold\n",
    "    df_select = current_match[(current_match['s2v_score'] > threshold)]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).index[0]\n",
    "        inchikey_match = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_match == inchikey_query)\n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_top_hit_s2v_top20.append(best_bet)\n",
    "\n",
    "\n",
    "#make arrays from lists\n",
    "test_top_hit_s2v_arr_top20 = np.array(test_top_hit_s2v_top20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 185,
   "metadata": {},
   "outputs": [],
   "source": [
    "#just look at mass + mod cosine\n",
    "#look at top20\n",
    "test_top_hit_mcos_top20 = []\n",
    "threshold = 0.4\n",
    "topn = 20 #only look at top 20\n",
    "\n",
    "for ID in range(len(documents_query_s2v)):\n",
    "    inchikey_query = documents_query_s2v[ID]._obj.get(\"inchikey\")[:14]\n",
    "    #select top20\n",
    "    current_match = found_matches_s2v[ID].sort_values(by=['mod_cosine_score'], ascending=False).iloc[:topn]\n",
    "\n",
    "    df_select = current_match[(current_match['mass_match'] == 1)\n",
    "                                  & (current_match['mod_cosine_score'] > threshold)]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['mod_cosine_score'], ascending=False).index[0]\n",
    "        inchikey_match = documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_match == inchikey_query)\n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_top_hit_mcos_top20.append(best_bet)\n",
    "\n",
    "\n",
    "#make arrays from lists\n",
    "test_top_hit_mcos_arr_top20 = np.array(test_top_hit_mcos_top20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 211,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Spectral library matching comparison 4 test sets looking at top 20\\n(>5 matching inchikey in library)')"
      ]
     },
     "execution_count": 211,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.style.use('ggplot')\n",
    "\n",
    "ylabels = ['1. mass + Spec2Vec',\n",
    "           '2. mass + Spec2Vec',\n",
    "           '3. mass + Spec2Vec',\n",
    "           '4. mass + Spec2Vec',\n",
    "           '1. mass + mod cosine',\n",
    "           '1. rf regression + mass change perc + pmass',\n",
    "           '2. rf regression + mass change perc + pmass',\n",
    "           '3. rf regression + mass change perc + pmass',\n",
    "           '4. rf regression + mass change perc + pmass',\n",
    "           '1. new (2) rf regression + mass change perc + pmass',\n",
    "           '2. new (2) rf regression + mass change perc + pmass',\n",
    "           '1. Spec2Vec + threshold',\n",
    "           '2. same mass + Spec2Vec']\n",
    "xlabels = ['true matches', 'false matches', 'no matches found']\n",
    "          #'mass + mod.cosine \\n + as backup: Spec2Vec + mod.cosine']\n",
    "data = np.array([[np.sum(test_m_s2v_arr_top20 == 1), np.sum(test_m_s2v_arr_top20 == 0), np.sum(test_m_s2v_arr_top20 == -1)],\n",
    "                 [np.sum(test_new_s2v_arr_top20 == 1), np.sum(test_new_s2v_arr_top20 == 0), np.sum(test_new_s2v_arr_top20 == -1)],\n",
    "                 [np.sum(test_new2_s2v_arr_top20 == 1), np.sum(test_new2_s2v_arr_top20 == 0), np.sum(test_new2_s2v_arr_top20 == -1)],\n",
    "                 [np.sum(test_new3_s2v_arr_top20 == 1), np.sum(test_new3_s2v_arr_top20 == 0), np.sum(test_new3_s2v_arr_top20 == -1)],\n",
    "                 [np.sum(test_top_hit_mcos_arr_top20 == 1), np.sum(test_top_hit_mcos_arr_top20 == 0), np.sum(test_top_hit_mcos_arr_top20 == -1)],\n",
    "                 [np.sum(test_m_rfreg_all_match_perc_pmass_arr_top20 == 1), np.sum(test_m_rfreg_all_match_perc_pmass_arr_top20 == 0),\n",
    "                 np.sum(test_m_rfreg_all_match_perc_pmass_arr_top20 == -1)],\n",
    "                 [np.sum(test_new_rfreg_all_match_perc_pmass_arr_top20 == 1), np.sum(test_new_rfreg_all_match_perc_pmass_arr_top20 == 0),\n",
    "                 np.sum(test_new_rfreg_all_match_perc_pmass_arr_top20 == -1)],\n",
    "                 [np.sum(test_new2_rfreg_all_match_perc_pmass_arr_top20 == 1), np.sum(test_new2_rfreg_all_match_perc_pmass_arr_top20 == 0),\n",
    "                 np.sum(test_new2_rfreg_all_match_perc_pmass_arr_top20 == -1)],\n",
    "                 [np.sum(test_new3_rfreg_all_match_perc_pmass_arr_top20 == 1), np.sum(test_new3_rfreg_all_match_perc_pmass_arr_top20 == 0),\n",
    "                 np.sum(test_new3_rfreg_all_match_perc_pmass_arr_top20 == -1)],\n",
    "                 [np.sum(test_old_rfreg_new_all_match_perc_pmass_arr_top20 == 1), np.sum(test_old_rfreg_new_all_match_perc_pmass_arr_top20 == 0),\n",
    "                 np.sum(test_old_rfreg_new_all_match_perc_pmass_arr_top20 == -1)],\n",
    "                 [np.sum(test_new_rfreg_new_all_match_perc_pmass_arr_top20 == 1), np.sum(test_new_rfreg_new_all_match_perc_pmass_arr_top20 == 0),\n",
    "                 np.sum(test_new_rfreg_new_all_match_perc_pmass_arr_top20 == -1)],\n",
    "                 [np.sum(test_top_hit_s2v_arr_top20 == 1), np.sum(test_top_hit_s2v_arr_top20 == 0), np.sum(test_top_hit_s2v_arr_top20 == -1)],\n",
    "                 [np.sum(test_new2_s2v_same_mass_arr_top20 == 1), np.sum(test_new2_s2v_same_mass_arr_top20 == 0), np.sum(test_new2_s2v_same_mass_arr_top20 == -1)]])\n",
    "\n",
    "dataframe = pd.DataFrame(data, columns=xlabels, index=ylabels)\n",
    "\n",
    "ax = dataframe.sort_values(by=['false matches'], ascending=True).plot.barh(stacked=True, edgecolor='none')\n",
    "\n",
    "horiz_offset = 1.03\n",
    "vert_offset = 1.\n",
    "ax.legend(bbox_to_anchor=(horiz_offset, vert_offset))\n",
    "plt.xlabel('Number of spectra')\n",
    "plt.title('Spectral library matching comparison 4 test sets looking at top 20\\n(>5 matching inchikey in library)')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 208,
   "metadata": {},
   "outputs": [],
   "source": [
    "#look at top20\n",
    "test_old_and_unique_s2v_top20 = []\n",
    "test_old_and_unique_mcos_top20 = []\n",
    "test_old_and_unique_s2v_same_mass_top20 = []\n",
    "test_old_and_unique_rfreg_all_match_perc_pmass_top20 = []\n",
    "threshold = 0.4\n",
    "topn = 20 #only look at top 20\n",
    "\n",
    "for ID in range(len(old_and_unique_documents_query_s2v)):\n",
    "    inchikey_query = old_and_unique_documents_query_s2v[ID]._obj.get(\"inchikey\")[:14]\n",
    "    #select top20\n",
    "    current_match = old_and_unique_found_matches_s2v[ID].sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "\n",
    "    # Scenario 2: mass + sort by Spec2Vec\n",
    "    df_select = current_match[(current_match['mass_match'] == 1)\n",
    "                                  & (current_match['s2v_score'] > threshold)]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).index[0]\n",
    "        inchikey_match = old_and_unique_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_match == inchikey_query)\n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_old_and_unique_s2v_top20.append(best_bet)\n",
    "    \n",
    "    # mass + sort by mod cosine\n",
    "    current_match_cos = old_and_unique_found_matches_s2v[ID].sort_values(by=['mod_cosine_score'], ascending=False).iloc[:topn]\n",
    "    df_select = current_match_cos[(current_match_cos['mass_match'] == 1)\n",
    "                                  & (current_match_cos['mod_cosine_score'] > threshold)]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['mod_cosine_score'], ascending=False).index[0]\n",
    "        inchikey_match = old_and_unique_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_match == inchikey_query)\n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_old_and_unique_mcos_top20.append(best_bet)\n",
    "\n",
    "    #RF regression all data new with perc mass change and parent mass\n",
    "    current_match_match_perc_pmass = old_and_unique_found_matches_s2v_match_perc_pmass_sim[ID].drop(['similarity','label'],axis=1).sort_values(by=['s2v_score'], ascending=False).iloc[:topn]\n",
    "    predictions = rfreg_all_new_match_perc_pmass[3].predict(current_match_match_perc_pmass)\n",
    "    #take the predicted match with highest s2v score\n",
    "    df_select = current_match_match_perc_pmass.iloc[predictions>=0.5]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).iloc[0].name\n",
    "        inchikey_selected = old_and_unique_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_selected == inchikey_query)        \n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_old_and_unique_rfreg_all_match_perc_pmass_top20.append(best_bet)\n",
    "    \n",
    "    # mass change 0 + sort by Spec2Vec\n",
    "    df_select = current_match_match_perc_pmass[(current_match_match_perc_pmass['perc_mass_change'] == 0)\n",
    "                                  & (current_match_match_perc_pmass['s2v_score'] > threshold)]\n",
    "    if df_select.shape[0] > 0:\n",
    "        best_match_ID = df_select.sort_values(by=['s2v_score'], ascending=False).index[0]\n",
    "        inchikey_match = old_and_unique_documents_library_s2v[best_match_ID]._obj.get(\"inchikey\")[:14]\n",
    "        best_bet = 1 * (inchikey_match == inchikey_query)\n",
    "    else:\n",
    "        best_bet = -1 # meaning: not found\n",
    "    test_old_and_unique_s2v_same_mass_top20.append(best_bet)\n",
    "\n",
    "#make arrays from lists\n",
    "test_old_and_unique_s2v_arr_top20 = np.array(test_old_and_unique_s2v_top20)\n",
    "test_old_and_unique_mcos_arr_top20 = np.array(test_old_and_unique_mcos_top20)\n",
    "test_old_and_unique_rfreg_all_match_perc_pmass_arr_top20 = np.array(test_old_and_unique_rfreg_all_match_perc_pmass_top20)\n",
    "test_old_and_unique_s2v_same_mass_arr_top20 = np.array(test_old_and_unique_s2v_same_mass_top20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 212,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Spectral library matching comparison looking at top 20\\n(50% of testset has a match in library)')"
      ]
     },
     "execution_count": 212,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.style.use('ggplot')\n",
    "\n",
    "ylabels = ['mass + Spec2Vec',\n",
    "           'mass + mod cosine',\n",
    "           'rf regression 2 + mass change perc + pmass',\n",
    "           'same mass + Spec2Vec']\n",
    "xlabels = ['true matches', 'false matches', 'no matches found']\n",
    "          #'mass + mod.cosine \\n + as backup: Spec2Vec + mod.cosine']\n",
    "data = np.array([[np.sum(test_old_and_unique_s2v_arr_top20 == 1), np.sum(test_old_and_unique_s2v_arr_top20 == 0), np.sum(test_old_and_unique_s2v_arr_top20 == -1)],\n",
    "                 [np.sum(test_old_and_unique_mcos_arr_top20 == 1), np.sum(test_old_and_unique_mcos_arr_top20 == 0), np.sum(test_old_and_unique_mcos_arr_top20 == -1)],\n",
    "                 [np.sum(test_old_and_unique_rfreg_all_match_perc_pmass_arr_top20 == 1), np.sum(test_old_and_unique_rfreg_all_match_perc_pmass_arr_top20 == 0),\n",
    "                 np.sum(test_old_and_unique_rfreg_all_match_perc_pmass_arr_top20 == -1)],\n",
    "                 [np.sum(test_old_and_unique_s2v_same_mass_arr_top20 == 1), np.sum(test_old_and_unique_s2v_same_mass_arr_top20 == 0),\n",
    "                  np.sum(test_old_and_unique_s2v_same_mass_arr_top20 == -1)]])\n",
    "\n",
    "dataframe = pd.DataFrame(data, columns=xlabels, index=ylabels)\n",
    "\n",
    "ax = dataframe.sort_values(by=['false matches'], ascending=True).plot.barh(stacked=True, edgecolor='none')\n",
    "\n",
    "horiz_offset = 1.03\n",
    "vert_offset = 1.\n",
    "ax.legend(bbox_to_anchor=(horiz_offset, vert_offset))\n",
    "plt.xlabel('Number of spectra')\n",
    "plt.title('Spectral library matching comparison looking at top 20\\n(50% of testset has a match in library)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training for similarity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "metadata": {},
   "outputs": [],
   "source": [
    "def convert_similarity_to_label(matches, cutoff = 0.6):\n",
    "    '''\n",
    "    Returns the same df with an extra column sim_label that indicates if similarity is above cutoff\n",
    "    \n",
    "    matches: list of pandas DataFrames, library matching result of query on library\n",
    "    cutoff: float, similarity cutoff for creating the new label\n",
    "    '''\n",
    "    matches_with_simlab = []\n",
    "    for match_old in matches:\n",
    "        match = match_old.copy()\n",
    "        sims = match['similarity']\n",
    "        simlabs = [1 if sim >= cutoff else 0 for sim in sims]\n",
    "        match['sim_label'] = simlabs\n",
    "        matches_with_simlab.append(match)\n",
    "    return matches_with_simlab\n",
    "found_matches_s2v_match_perc_pmass_simlabel = convert_similarity_to_label(found_matches_s2v_match_perc_pmass_sim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 173,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>label</th>\n",
       "      <th>perc_mass_change</th>\n",
       "      <th>parent_mass</th>\n",
       "      <th>similarity</th>\n",
       "      <th>sim_label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>46003</th>\n",
       "      <td>0.960780</td>\n",
       "      <td>3</td>\n",
       "      <td>0.960780</td>\n",
       "      <td>3</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.887800</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>508.340724</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38195</th>\n",
       "      <td>0.049197</td>\n",
       "      <td>1</td>\n",
       "      <td>0.971344</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.858902</td>\n",
       "      <td>0</td>\n",
       "      <td>2.757206</td>\n",
       "      <td>522.356724</td>\n",
       "      <td>0.774278</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48062</th>\n",
       "      <td>0.043040</td>\n",
       "      <td>2</td>\n",
       "      <td>0.043040</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.848561</td>\n",
       "      <td>0</td>\n",
       "      <td>1.567649</td>\n",
       "      <td>516.309724</td>\n",
       "      <td>0.907563</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45969</th>\n",
       "      <td>0.049498</td>\n",
       "      <td>1</td>\n",
       "      <td>0.961919</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.844523</td>\n",
       "      <td>0</td>\n",
       "      <td>8.271421</td>\n",
       "      <td>466.293724</td>\n",
       "      <td>0.774278</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32755</th>\n",
       "      <td>0.119678</td>\n",
       "      <td>1</td>\n",
       "      <td>0.608107</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.811812</td>\n",
       "      <td>0</td>\n",
       "      <td>2.757206</td>\n",
       "      <td>522.356724</td>\n",
       "      <td>0.774278</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33654</th>\n",
       "      <td>0.944742</td>\n",
       "      <td>2</td>\n",
       "      <td>0.944742</td>\n",
       "      <td>2</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.799576</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>508.340724</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33166</th>\n",
       "      <td>0.049271</td>\n",
       "      <td>1</td>\n",
       "      <td>0.949622</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.790485</td>\n",
       "      <td>0</td>\n",
       "      <td>8.271421</td>\n",
       "      <td>466.293724</td>\n",
       "      <td>0.774278</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30807</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.961388</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.760594</td>\n",
       "      <td>0</td>\n",
       "      <td>9.083475</td>\n",
       "      <td>554.515724</td>\n",
       "      <td>0.170132</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>56270</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.961388</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.747925</td>\n",
       "      <td>0</td>\n",
       "      <td>25.626316</td>\n",
       "      <td>638.609724</td>\n",
       "      <td>0.170132</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34114</th>\n",
       "      <td>0.046938</td>\n",
       "      <td>1</td>\n",
       "      <td>0.979920</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.739805</td>\n",
       "      <td>0</td>\n",
       "      <td>2.757206</td>\n",
       "      <td>494.324724</td>\n",
       "      <td>0.774278</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>56863</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.961388</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.735438</td>\n",
       "      <td>0</td>\n",
       "      <td>31.140531</td>\n",
       "      <td>666.640724</td>\n",
       "      <td>0.170132</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45979</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.938445</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.732793</td>\n",
       "      <td>0</td>\n",
       "      <td>3.153593</td>\n",
       "      <td>492.309724</td>\n",
       "      <td>0.944606</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>53958</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.730020</td>\n",
       "      <td>0</td>\n",
       "      <td>27.994019</td>\n",
       "      <td>650.645724</td>\n",
       "      <td>0.175758</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>57799</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.961388</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.728967</td>\n",
       "      <td>0</td>\n",
       "      <td>22.479804</td>\n",
       "      <td>622.614724</td>\n",
       "      <td>0.175758</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33190</th>\n",
       "      <td>0.045540</td>\n",
       "      <td>1</td>\n",
       "      <td>0.982362</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.727989</td>\n",
       "      <td>0</td>\n",
       "      <td>2.757206</td>\n",
       "      <td>494.324724</td>\n",
       "      <td>0.774278</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38480</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.961388</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.726993</td>\n",
       "      <td>0</td>\n",
       "      <td>27.994019</td>\n",
       "      <td>650.645724</td>\n",
       "      <td>0.175758</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38277</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.961388</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.724805</td>\n",
       "      <td>0</td>\n",
       "      <td>5.936963</td>\n",
       "      <td>538.520724</td>\n",
       "      <td>0.175758</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32721</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.961388</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.724702</td>\n",
       "      <td>0</td>\n",
       "      <td>11.406719</td>\n",
       "      <td>566.325724</td>\n",
       "      <td>0.797844</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>56286</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.961388</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.724587</td>\n",
       "      <td>0</td>\n",
       "      <td>31.140531</td>\n",
       "      <td>666.640724</td>\n",
       "      <td>0.170132</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45044</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.961388</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.721622</td>\n",
       "      <td>0</td>\n",
       "      <td>22.479804</td>\n",
       "      <td>622.614724</td>\n",
       "      <td>0.175758</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "46003      0.960780               3          0.960780                   3   \n",
       "38195      0.049197               1          0.971344                   3   \n",
       "48062      0.043040               2          0.043040                   2   \n",
       "45969      0.049498               1          0.961919                   2   \n",
       "32755      0.119678               1          0.608107                   2   \n",
       "33654      0.944742               2          0.944742                   2   \n",
       "33166      0.049271               1          0.949622                   2   \n",
       "30807      0.000000               0          0.961388                   1   \n",
       "56270      0.000000               0          0.961388                   1   \n",
       "34114      0.046938               1          0.979920                   2   \n",
       "56863      0.000000               0          0.961388                   1   \n",
       "45979      0.000000               0          0.938445                   3   \n",
       "53958      0.000000               0          0.000000                   0   \n",
       "57799      0.000000               0          0.961388                   1   \n",
       "33190      0.045540               1          0.982362                   3   \n",
       "38480      0.000000               0          0.961388                   1   \n",
       "38277      0.000000               0          0.961388                   1   \n",
       "32721      0.000000               0          0.961388                   1   \n",
       "56286      0.000000               0          0.961388                   1   \n",
       "45044      0.000000               0          0.961388                   1   \n",
       "\n",
       "       mass_match  s2v_score  label  perc_mass_change  parent_mass  \\\n",
       "46003         1.0   0.887800      1          0.000000   508.340724   \n",
       "38195         0.0   0.858902      0          2.757206   522.356724   \n",
       "48062         0.0   0.848561      0          1.567649   516.309724   \n",
       "45969         0.0   0.844523      0          8.271421   466.293724   \n",
       "32755         0.0   0.811812      0          2.757206   522.356724   \n",
       "33654         1.0   0.799576      1          0.000000   508.340724   \n",
       "33166         0.0   0.790485      0          8.271421   466.293724   \n",
       "30807         0.0   0.760594      0          9.083475   554.515724   \n",
       "56270         0.0   0.747925      0         25.626316   638.609724   \n",
       "34114         0.0   0.739805      0          2.757206   494.324724   \n",
       "56863         0.0   0.735438      0         31.140531   666.640724   \n",
       "45979         0.0   0.732793      0          3.153593   492.309724   \n",
       "53958         0.0   0.730020      0         27.994019   650.645724   \n",
       "57799         0.0   0.728967      0         22.479804   622.614724   \n",
       "33190         0.0   0.727989      0          2.757206   494.324724   \n",
       "38480         0.0   0.726993      0         27.994019   650.645724   \n",
       "38277         0.0   0.724805      0          5.936963   538.520724   \n",
       "32721         0.0   0.724702      0         11.406719   566.325724   \n",
       "56286         0.0   0.724587      0         31.140531   666.640724   \n",
       "45044         0.0   0.721622      0         22.479804   622.614724   \n",
       "\n",
       "       similarity  sim_label  \n",
       "46003    1.000000          1  \n",
       "38195    0.774278          1  \n",
       "48062    0.907563          1  \n",
       "45969    0.774278          1  \n",
       "32755    0.774278          1  \n",
       "33654    1.000000          1  \n",
       "33166    0.774278          1  \n",
       "30807    0.170132          0  \n",
       "56270    0.170132          0  \n",
       "34114    0.774278          1  \n",
       "56863    0.170132          0  \n",
       "45979    0.944606          1  \n",
       "53958    0.175758          0  \n",
       "57799    0.175758          0  \n",
       "33190    0.774278          1  \n",
       "38480    0.175758          0  \n",
       "38277    0.175758          0  \n",
       "32721    0.797844          1  \n",
       "56286    0.170132          0  \n",
       "45044    0.175758          0  "
      ]
     },
     "execution_count": 173,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "found_matches_s2v_match_perc_pmass_simlabel[0].sort_values(by=['s2v_score'],ascending=False).iloc[:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 174,
   "metadata": {},
   "outputs": [],
   "source": [
    "all_found_matches_s2v_match_perc_pmass_simlabel = found_matches_s2v_match_perc_pmass_simlabel[0].append(found_matches_s2v_match_perc_pmass_simlabel[1:])\n",
    "\n",
    "all_found_matches_s2v_match_perc_pmass_simlabel_drop = all_found_matches_s2v_match_perc_pmass_simlabel.drop(['label','similarity'],axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 175,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>perc_mass_change</th>\n",
       "      <th>parent_mass</th>\n",
       "      <th>sim_label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>65637</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>11</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>11</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>191.130724</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40198</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>37</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>37</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>789.697724</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40059</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>80</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>80</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>832.585724</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>69565</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>15</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>15</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>844.305724</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66426</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>54</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>54</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>202.106724</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1265</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.243909</td>\n",
       "      <td>0.002615</td>\n",
       "      <td>344.116724</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30366</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.245904</td>\n",
       "      <td>0.145163</td>\n",
       "      <td>567.499724</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38243</th>\n",
       "      <td>0.000005</td>\n",
       "      <td>1</td>\n",
       "      <td>0.000005</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.265501</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>730.539724</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32419</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.265508</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>730.539724</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>90987</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.275764</td>\n",
       "      <td>0.020131</td>\n",
       "      <td>422.142724</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>239233 rows × 9 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "65637      1.000000              11          1.000000                  11   \n",
       "40198      1.000000              37          1.000000                  37   \n",
       "40059      1.000000              80          1.000000                  80   \n",
       "69565      1.000000              15          1.000000                  15   \n",
       "66426      1.000000              54          1.000000                  54   \n",
       "...             ...             ...               ...                 ...   \n",
       "1265       0.000000               0          0.000000                   0   \n",
       "30366      0.000000               0          0.000000                   0   \n",
       "38243      0.000005               1          0.000005                   1   \n",
       "32419      0.000000               0          0.000000                   0   \n",
       "90987      0.000000               0          0.000000                   0   \n",
       "\n",
       "       mass_match  s2v_score  perc_mass_change  parent_mass  sim_label  \n",
       "65637         1.0   1.000000          0.000000   191.130724          1  \n",
       "40198         1.0   1.000000          0.000000   789.697724          1  \n",
       "40059         1.0   1.000000          0.000000   832.585724          1  \n",
       "69565         1.0   1.000000          0.000000   844.305724          1  \n",
       "66426         1.0   1.000000          0.000000   202.106724          1  \n",
       "...           ...        ...               ...          ...        ...  \n",
       "1265          1.0  -0.243909          0.002615   344.116724          0  \n",
       "30366         1.0  -0.245904          0.145163   567.499724          0  \n",
       "38243         1.0  -0.265501          0.000000   730.539724          1  \n",
       "32419         1.0  -0.265508          0.000000   730.539724          1  \n",
       "90987         1.0  -0.275764          0.020131   422.142724          0  \n",
       "\n",
       "[239233 rows x 9 columns]"
      ]
     },
     "execution_count": 175,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_found_matches_s2v_match_perc_pmass_simlabel_drop.sort_values('s2v_score',ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 176,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label distribution: {0: 216573, 1: 22660}\n",
      "Training Features Shape: (179424, 8)\n",
      "Training Labels Shape: (179424,)\n",
      "Training label distribution: {0: 162428, 1: 16996}\n",
      "Testing Features Shape: (59809, 8)\n",
      "Testing Labels Shape: (59809,)\n",
      "Test label distribution: {0: 54145, 1: 5664}\n",
      "\n",
      "Predicted label distribution: {0.0: 49518, 0.1: 2621, 0.2: 1130, 0.3: 622, 0.4: 569, 0.5: 430, 0.6: 456, 0.7: 404, 0.8: 562, 0.9: 710, 1.0: 2787}\n",
      "Predictions: \n",
      "0->0 1->1\n",
      "53656 4650\n",
      "0->1 1->0\n",
      "489 1014\n",
      "Accuracy: 0.97\n",
      "\n",
      "Variable: perc_mass_change     Importance: 0.48\n",
      "Variable: parent_mass          Importance: 0.19\n",
      "Variable: s2v_score            Importance: 0.18\n",
      "Variable: mod_cosine_score     Importance: 0.05\n",
      "Variable: cosine_score         Importance: 0.04\n",
      "Variable: cosine_matches       Importance: 0.03\n",
      "Variable: mod_cosine_matches   Importance: 0.03\n",
      "Variable: mass_match           Importance: 0.0\n"
     ]
    }
   ],
   "source": [
    "rfreg_all_match_perc_pmass_simlabel = do_random_forest(all_found_matches_s2v_match_perc_pmass_simlabel_drop,\n",
    "                                                       'sim_label', rf_regr = True,  verbose = True, num_est = 200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 177,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>perc_mass_change</th>\n",
       "      <th>parent_mass</th>\n",
       "      <th>sim_label</th>\n",
       "      <th>prediction</th>\n",
       "      <th>correct_guess</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>35122</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>15.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>15.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>844.305724</td>\n",
       "      <td>1</td>\n",
       "      <td>1.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34791</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>41.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>41.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>203.130724</td>\n",
       "      <td>1</td>\n",
       "      <td>0.845</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40735</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>123.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>123.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>451.185724</td>\n",
       "      <td>1</td>\n",
       "      <td>0.970</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1733</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>16.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>16.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>212.025724</td>\n",
       "      <td>1</td>\n",
       "      <td>0.645</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40191</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>55.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>55.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>199.077724</td>\n",
       "      <td>1</td>\n",
       "      <td>0.845</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25388</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>49.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>49.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>913.822724</td>\n",
       "      <td>1</td>\n",
       "      <td>1.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44288</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>34.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>34.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>242.094724</td>\n",
       "      <td>1</td>\n",
       "      <td>0.705</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26903</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>24.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>24.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>532.303724</td>\n",
       "      <td>1</td>\n",
       "      <td>1.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49212</th>\n",
       "      <td>0.999343</td>\n",
       "      <td>16.0</td>\n",
       "      <td>0.999343</td>\n",
       "      <td>16.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.999699</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>175.048724</td>\n",
       "      <td>1</td>\n",
       "      <td>0.920</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25597</th>\n",
       "      <td>0.999543</td>\n",
       "      <td>13.0</td>\n",
       "      <td>0.999543</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.998715</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>320.209724</td>\n",
       "      <td>1</td>\n",
       "      <td>1.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16448</th>\n",
       "      <td>0.999735</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.999753</td>\n",
       "      <td>68.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.998407</td>\n",
       "      <td>0.246150</td>\n",
       "      <td>816.591724</td>\n",
       "      <td>1</td>\n",
       "      <td>0.800</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32228</th>\n",
       "      <td>0.999806</td>\n",
       "      <td>61.0</td>\n",
       "      <td>0.999808</td>\n",
       "      <td>62.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.998219</td>\n",
       "      <td>7.154190</td>\n",
       "      <td>727.575724</td>\n",
       "      <td>1</td>\n",
       "      <td>0.635</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36591</th>\n",
       "      <td>0.999870</td>\n",
       "      <td>70.0</td>\n",
       "      <td>0.999896</td>\n",
       "      <td>71.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.998122</td>\n",
       "      <td>6.507486</td>\n",
       "      <td>718.575724</td>\n",
       "      <td>1</td>\n",
       "      <td>0.780</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4482</th>\n",
       "      <td>0.999829</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.999829</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997947</td>\n",
       "      <td>1.110912</td>\n",
       "      <td>827.700724</td>\n",
       "      <td>0</td>\n",
       "      <td>0.800</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10890</th>\n",
       "      <td>0.999955</td>\n",
       "      <td>72.0</td>\n",
       "      <td>0.999972</td>\n",
       "      <td>73.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997897</td>\n",
       "      <td>0.262298</td>\n",
       "      <td>766.575724</td>\n",
       "      <td>1</td>\n",
       "      <td>0.885</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31930</th>\n",
       "      <td>0.999890</td>\n",
       "      <td>66.0</td>\n",
       "      <td>0.999892</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997840</td>\n",
       "      <td>7.411323</td>\n",
       "      <td>725.560724</td>\n",
       "      <td>1</td>\n",
       "      <td>0.760</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21224</th>\n",
       "      <td>0.999842</td>\n",
       "      <td>59.0</td>\n",
       "      <td>0.999842</td>\n",
       "      <td>59.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997809</td>\n",
       "      <td>3.699863</td>\n",
       "      <td>729.591724</td>\n",
       "      <td>1</td>\n",
       "      <td>0.780</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>53364</th>\n",
       "      <td>0.999891</td>\n",
       "      <td>61.0</td>\n",
       "      <td>0.999891</td>\n",
       "      <td>61.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.997776</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>757.622724</td>\n",
       "      <td>1</td>\n",
       "      <td>1.000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8183</th>\n",
       "      <td>0.999908</td>\n",
       "      <td>74.0</td>\n",
       "      <td>0.999908</td>\n",
       "      <td>74.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997749</td>\n",
       "      <td>10.544870</td>\n",
       "      <td>687.544724</td>\n",
       "      <td>0</td>\n",
       "      <td>0.570</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27328</th>\n",
       "      <td>0.999741</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.999741</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.997732</td>\n",
       "      <td>1.110912</td>\n",
       "      <td>827.700724</td>\n",
       "      <td>0</td>\n",
       "      <td>0.785</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "35122      1.000000            15.0          1.000000                15.0   \n",
       "34791      1.000000            41.0          1.000000                41.0   \n",
       "40735      1.000000           123.0          1.000000               123.0   \n",
       "1733       1.000000            16.0          1.000000                16.0   \n",
       "40191      1.000000            55.0          1.000000                55.0   \n",
       "25388      1.000000            49.0          1.000000                49.0   \n",
       "44288      1.000000            34.0          1.000000                34.0   \n",
       "26903      1.000000            24.0          1.000000                24.0   \n",
       "49212      0.999343            16.0          0.999343                16.0   \n",
       "25597      0.999543            13.0          0.999543                13.0   \n",
       "16448      0.999735            67.0          0.999753                68.0   \n",
       "32228      0.999806            61.0          0.999808                62.0   \n",
       "36591      0.999870            70.0          0.999896                71.0   \n",
       "4482       0.999829            67.0          0.999829                67.0   \n",
       "10890      0.999955            72.0          0.999972                73.0   \n",
       "31930      0.999890            66.0          0.999892                67.0   \n",
       "21224      0.999842            59.0          0.999842                59.0   \n",
       "53364      0.999891            61.0          0.999891                61.0   \n",
       "8183       0.999908            74.0          0.999908                74.0   \n",
       "27328      0.999741            67.0          0.999741                67.0   \n",
       "\n",
       "       mass_match  s2v_score  perc_mass_change  parent_mass  sim_label  \\\n",
       "35122         1.0   1.000000          0.000000   844.305724          1   \n",
       "34791         1.0   1.000000          0.000000   203.130724          1   \n",
       "40735         1.0   1.000000          0.000000   451.185724          1   \n",
       "1733          1.0   1.000000          0.000000   212.025724          1   \n",
       "40191         1.0   1.000000          0.000000   199.077724          1   \n",
       "25388         1.0   1.000000          0.000000   913.822724          1   \n",
       "44288         1.0   1.000000          0.000000   242.094724          1   \n",
       "26903         1.0   1.000000          0.000000   532.303724          1   \n",
       "49212         1.0   0.999699          0.000000   175.048724          1   \n",
       "25597         1.0   0.998715          0.000000   320.209724          1   \n",
       "16448         0.0   0.998407          0.246150   816.591724          1   \n",
       "32228         0.0   0.998219          7.154190   727.575724          1   \n",
       "36591         0.0   0.998122          6.507486   718.575724          1   \n",
       "4482          0.0   0.997947          1.110912   827.700724          0   \n",
       "10890         0.0   0.997897          0.262298   766.575724          1   \n",
       "31930         0.0   0.997840          7.411323   725.560724          1   \n",
       "21224         0.0   0.997809          3.699863   729.591724          1   \n",
       "53364         1.0   0.997776          0.000000   757.622724          1   \n",
       "8183          0.0   0.997749         10.544870   687.544724          0   \n",
       "27328         0.0   0.997732          1.110912   827.700724          0   \n",
       "\n",
       "       prediction  correct_guess  \n",
       "35122       1.000           True  \n",
       "34791       0.845           True  \n",
       "40735       0.970           True  \n",
       "1733        0.645           True  \n",
       "40191       0.845           True  \n",
       "25388       1.000           True  \n",
       "44288       0.705           True  \n",
       "26903       1.000           True  \n",
       "49212       0.920           True  \n",
       "25597       1.000           True  \n",
       "16448       0.800           True  \n",
       "32228       0.635           True  \n",
       "36591       0.780           True  \n",
       "4482        0.800          False  \n",
       "10890       0.885           True  \n",
       "31930       0.760           True  \n",
       "21224       0.780           True  \n",
       "53364       1.000           True  \n",
       "8183        0.570          False  \n",
       "27328       0.785          False  "
      ]
     },
     "execution_count": 177,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#visualising performance on the test set\n",
    "test_set_rfreg_all_match_perc_pmass_simlabel = pd.DataFrame(rfreg_all_match_perc_pmass_simlabel[0], columns = cols+['perc_mass_change','parent_mass'])\n",
    "test_set_rfreg_all_match_perc_pmass_simlabel['sim_label'] = rfreg_all_match_perc_pmass_simlabel[1]\n",
    "test_set_rfreg_all_match_perc_pmass_simlabel['prediction'] = rfreg_all_match_perc_pmass_simlabel[2]\n",
    "test_set_rfreg_all_match_perc_pmass_simlabel['correct_guess'] = (round(test_set_rfreg_all_match_perc_pmass_simlabel['sim_label']) == round(test_set_rfreg_all_match_perc_pmass_simlabel['prediction']))\n",
    "test_set_rfreg_all_match_perc_pmass_simlabel.sort_values(by=['s2v_score'],ascending=False).iloc[:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 178,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>perc_mass_change</th>\n",
       "      <th>parent_mass</th>\n",
       "      <th>sim_label</th>\n",
       "      <th>prediction</th>\n",
       "      <th>correct_guess</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.001929</td>\n",
       "      <td>9.0</td>\n",
       "      <td>0.001929</td>\n",
       "      <td>9.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.069848</td>\n",
       "      <td>0.012843</td>\n",
       "      <td>358.211724</td>\n",
       "      <td>0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.003362</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.004176</td>\n",
       "      <td>24.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.213126</td>\n",
       "      <td>0.008842</td>\n",
       "      <td>260.104724</td>\n",
       "      <td>0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.003422</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.047706</td>\n",
       "      <td>0.476897</td>\n",
       "      <td>194.080724</td>\n",
       "      <td>0</td>\n",
       "      <td>0.01</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.003400</td>\n",
       "      <td>0.029544</td>\n",
       "      <td>494.324724</td>\n",
       "      <td>0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.000478</td>\n",
       "      <td>7.0</td>\n",
       "      <td>0.000478</td>\n",
       "      <td>7.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.026660</td>\n",
       "      <td>0.067087</td>\n",
       "      <td>392.292724</td>\n",
       "      <td>0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>59804</th>\n",
       "      <td>0.003739</td>\n",
       "      <td>13.0</td>\n",
       "      <td>0.004599</td>\n",
       "      <td>15.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.048225</td>\n",
       "      <td>0.083488</td>\n",
       "      <td>861.676724</td>\n",
       "      <td>0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>59805</th>\n",
       "      <td>0.000933</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.019546</td>\n",
       "      <td>4.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.086335</td>\n",
       "      <td>0.440943</td>\n",
       "      <td>220.041724</td>\n",
       "      <td>0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>59806</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.000484</td>\n",
       "      <td>0.094323</td>\n",
       "      <td>317.293724</td>\n",
       "      <td>0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>59807</th>\n",
       "      <td>0.985551</td>\n",
       "      <td>20.0</td>\n",
       "      <td>0.985551</td>\n",
       "      <td>20.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.856380</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>550.095724</td>\n",
       "      <td>1</td>\n",
       "      <td>1.00</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>59808</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-0.053099</td>\n",
       "      <td>0.108389</td>\n",
       "      <td>593.514724</td>\n",
       "      <td>0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>59809 rows × 11 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "0          0.001929             9.0          0.001929                 9.0   \n",
       "1          0.003362            15.0          0.004176                24.0   \n",
       "2          0.000000             0.0          0.003422                 1.0   \n",
       "3          0.000000             0.0          0.000000                 0.0   \n",
       "4          0.000478             7.0          0.000478                 7.0   \n",
       "...             ...             ...               ...                 ...   \n",
       "59804      0.003739            13.0          0.004599                15.0   \n",
       "59805      0.000933             2.0          0.019546                 4.0   \n",
       "59806      0.000000             0.0          0.000000                 0.0   \n",
       "59807      0.985551            20.0          0.985551                20.0   \n",
       "59808      0.000000             0.0          0.000000                 0.0   \n",
       "\n",
       "       mass_match  s2v_score  perc_mass_change  parent_mass  sim_label  \\\n",
       "0             1.0   0.069848          0.012843   358.211724          0   \n",
       "1             1.0   0.213126          0.008842   260.104724          0   \n",
       "2             1.0   0.047706          0.476897   194.080724          0   \n",
       "3             1.0   0.003400          0.029544   494.324724          0   \n",
       "4             1.0   0.026660          0.067087   392.292724          0   \n",
       "...           ...        ...               ...          ...        ...   \n",
       "59804         1.0   0.048225          0.083488   861.676724          0   \n",
       "59805         1.0   0.086335          0.440943   220.041724          0   \n",
       "59806         1.0   0.000484          0.094323   317.293724          0   \n",
       "59807         1.0   0.856380          0.000000   550.095724          1   \n",
       "59808         1.0  -0.053099          0.108389   593.514724          0   \n",
       "\n",
       "       prediction  correct_guess  \n",
       "0            0.00           True  \n",
       "1            0.00           True  \n",
       "2            0.01           True  \n",
       "3            0.00           True  \n",
       "4            0.00           True  \n",
       "...           ...            ...  \n",
       "59804        0.00           True  \n",
       "59805        0.00           True  \n",
       "59806        0.00           True  \n",
       "59807        1.00           True  \n",
       "59808        0.00           True  \n",
       "\n",
       "[59809 rows x 11 columns]"
      ]
     },
     "execution_count": 178,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#try to see performance for each bin of prediction label (bin 0.7 -> correctly predicted 60%, uncorrect 40% something like this)\n",
    "\n",
    "test_set_rfreg_all_match_perc_pmass_simlabel"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Investigate missing mass matches for matching labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 179,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "18 False\n",
      "25 False\n",
      "29 False\n",
      "30 False\n",
      "32 False\n",
      "33 True\n",
      "44 False\n",
      "48 False\n",
      "50 False\n",
      "56 False\n",
      "75 False\n",
      "81 False\n",
      "82 False\n",
      "91 False\n",
      "93 False\n",
      "94 False\n",
      "101 True\n",
      "103 False\n",
      "113 True\n",
      "124 False\n",
      "126 False\n",
      "152 False\n",
      "155 False\n",
      "158 False\n",
      "171 True\n",
      "173 False\n",
      "183 False\n",
      "186 False\n",
      "196 False\n",
      "204 False\n",
      "214 False\n",
      "215 False\n",
      "222 False\n",
      "224 False\n",
      "227 False\n",
      "231 True\n",
      "232 True\n",
      "233 False\n",
      "237 False\n",
      "249 False\n",
      "275 False\n",
      "279 False\n",
      "280 False\n",
      "283 False\n",
      "287 False\n",
      "292 False\n",
      "297 False\n",
      "298 False\n",
      "300 False\n",
      "303 False\n",
      "304 True\n",
      "307 False\n",
      "310 False\n",
      "315 False\n",
      "336 False\n",
      "337 False\n",
      "338 False\n",
      "343 False\n",
      "364 False\n",
      "366 False\n",
      "373 False\n",
      "388 False\n",
      "389 False\n",
      "393 False\n",
      "401 False\n",
      "406 False\n",
      "412 False\n",
      "417 False\n",
      "419 False\n",
      "420 False\n",
      "427 False\n",
      "430 False\n",
      "449 True\n",
      "455 False\n",
      "464 False\n",
      "473 False\n",
      "531 False\n",
      "540 False\n",
      "543 True\n",
      "556 False\n",
      "568 False\n",
      "572 True\n",
      "591 False\n",
      "593 False\n",
      "597 False\n",
      "604 False\n",
      "613 False\n",
      "623 False\n",
      "625 False\n",
      "631 False\n",
      "635 False\n",
      "636 False\n",
      "640 False\n",
      "641 False\n",
      "649 False\n",
      "657 False\n",
      "659 False\n",
      "661 False\n",
      "662 False\n",
      "663 True\n",
      "665 False\n",
      "666 False\n",
      "676 False\n",
      "677 False\n",
      "685 False\n",
      "688 False\n",
      "693 False\n",
      "707 False\n",
      "711 False\n",
      "715 True\n",
      "718 False\n",
      "723 False\n",
      "730 True\n",
      "739 False\n",
      "742 False\n",
      "745 True\n",
      "753 False\n",
      "756 False\n",
      "773 False\n",
      "774 False\n",
      "779 False\n",
      "784 False\n",
      "800 False\n",
      "801 False\n",
      "808 False\n",
      "811 False\n",
      "822 False\n",
      "823 False\n",
      "831 True\n",
      "832 False\n",
      "833 False\n",
      "845 True\n",
      "847 False\n",
      "867 False\n",
      "876 False\n",
      "888 False\n",
      "894 False\n",
      "896 False\n",
      "910 False\n",
      "917 True\n",
      "918 False\n",
      "920 False\n",
      "924 False\n",
      "943 False\n",
      "945 False\n",
      "947 True\n",
      "955 False\n",
      "963 False\n",
      "974 True\n",
      "981 False\n",
      "983 False\n",
      "992 False\n",
      "995 False\n",
      "998 False\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "154"
      ]
     },
     "execution_count": 179,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#find a spectrum with true matches but without mass_matches\n",
    "all_trues = []\n",
    "inds = []\n",
    "for ind in range(len(found_matches_s2v)):\n",
    "    \n",
    "    predi = all_result[3].predict(found_matches_s2v[ind])\n",
    "    dfi = found_matches_s2v_labels[ind].copy()\n",
    "    dfi['prediction'] = predi\n",
    "    sorts2v = dfi.sort_values(by=['s2v_score'], ascending=False).iloc[:20]\n",
    "    select = sorts2v[(sorts2v['label']==1) & (sorts2v['prediction']==0) & (sorts2v['mass_match']==0)]\n",
    "    if select.shape[0] > 0:\n",
    "        all_true = all(np.array(sorts2v.iloc[0])==np.array(select.iloc[0]))\n",
    "        print(ind, all_true) #if top select is top match\n",
    "        inds.append(ind)\n",
    "        if all_true:\n",
    "            all_trues.append(ind)\n",
    "len(all_trues)\n",
    "len(inds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 180,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cosine_score</th>\n",
       "      <th>cosine_matches</th>\n",
       "      <th>mod_cosine_score</th>\n",
       "      <th>mod_cosine_matches</th>\n",
       "      <th>mass_match</th>\n",
       "      <th>s2v_score</th>\n",
       "      <th>label</th>\n",
       "      <th>prediction</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>66633</th>\n",
       "      <td>0.664616</td>\n",
       "      <td>45</td>\n",
       "      <td>0.750967</td>\n",
       "      <td>50</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.825818</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66795</th>\n",
       "      <td>0.608122</td>\n",
       "      <td>39</td>\n",
       "      <td>0.639218</td>\n",
       "      <td>43</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.770255</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66678</th>\n",
       "      <td>0.094834</td>\n",
       "      <td>38</td>\n",
       "      <td>0.249452</td>\n",
       "      <td>45</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.697615</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>64335</th>\n",
       "      <td>0.023554</td>\n",
       "      <td>37</td>\n",
       "      <td>0.058843</td>\n",
       "      <td>48</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.654275</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66152</th>\n",
       "      <td>0.023938</td>\n",
       "      <td>32</td>\n",
       "      <td>0.028580</td>\n",
       "      <td>60</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.653601</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66722</th>\n",
       "      <td>0.023938</td>\n",
       "      <td>32</td>\n",
       "      <td>0.028580</td>\n",
       "      <td>60</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.653601</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66744</th>\n",
       "      <td>0.049016</td>\n",
       "      <td>27</td>\n",
       "      <td>0.050029</td>\n",
       "      <td>38</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.651371</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>64311</th>\n",
       "      <td>0.049016</td>\n",
       "      <td>27</td>\n",
       "      <td>0.050029</td>\n",
       "      <td>38</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.651371</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66727</th>\n",
       "      <td>0.005148</td>\n",
       "      <td>16</td>\n",
       "      <td>0.119018</td>\n",
       "      <td>28</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.639592</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>68782</th>\n",
       "      <td>0.245597</td>\n",
       "      <td>18</td>\n",
       "      <td>0.346483</td>\n",
       "      <td>23</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.635570</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>62839</th>\n",
       "      <td>0.003691</td>\n",
       "      <td>11</td>\n",
       "      <td>0.035250</td>\n",
       "      <td>26</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.632343</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>64229</th>\n",
       "      <td>0.007235</td>\n",
       "      <td>13</td>\n",
       "      <td>0.306120</td>\n",
       "      <td>26</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.626987</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>62566</th>\n",
       "      <td>0.000207</td>\n",
       "      <td>5</td>\n",
       "      <td>0.226556</td>\n",
       "      <td>14</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.624334</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>64271</th>\n",
       "      <td>0.042544</td>\n",
       "      <td>28</td>\n",
       "      <td>0.047961</td>\n",
       "      <td>40</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.609262</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>68768</th>\n",
       "      <td>0.042544</td>\n",
       "      <td>28</td>\n",
       "      <td>0.047961</td>\n",
       "      <td>40</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.608923</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>63474</th>\n",
       "      <td>0.000081</td>\n",
       "      <td>2</td>\n",
       "      <td>0.062082</td>\n",
       "      <td>13</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.607861</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>63569</th>\n",
       "      <td>0.000405</td>\n",
       "      <td>2</td>\n",
       "      <td>0.014027</td>\n",
       "      <td>13</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.607152</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>63564</th>\n",
       "      <td>0.000181</td>\n",
       "      <td>3</td>\n",
       "      <td>0.007369</td>\n",
       "      <td>7</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.592299</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>64097</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.009182</td>\n",
       "      <td>10</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.590802</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>63690</th>\n",
       "      <td>0.000307</td>\n",
       "      <td>2</td>\n",
       "      <td>0.011019</td>\n",
       "      <td>6</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.586521</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       cosine_score  cosine_matches  mod_cosine_score  mod_cosine_matches  \\\n",
       "66633      0.664616              45          0.750967                  50   \n",
       "66795      0.608122              39          0.639218                  43   \n",
       "66678      0.094834              38          0.249452                  45   \n",
       "64335      0.023554              37          0.058843                  48   \n",
       "66152      0.023938              32          0.028580                  60   \n",
       "66722      0.023938              32          0.028580                  60   \n",
       "66744      0.049016              27          0.050029                  38   \n",
       "64311      0.049016              27          0.050029                  38   \n",
       "66727      0.005148              16          0.119018                  28   \n",
       "68782      0.245597              18          0.346483                  23   \n",
       "62839      0.003691              11          0.035250                  26   \n",
       "64229      0.007235              13          0.306120                  26   \n",
       "62566      0.000207               5          0.226556                  14   \n",
       "64271      0.042544              28          0.047961                  40   \n",
       "68768      0.042544              28          0.047961                  40   \n",
       "63474      0.000081               2          0.062082                  13   \n",
       "63569      0.000405               2          0.014027                  13   \n",
       "63564      0.000181               3          0.007369                   7   \n",
       "64097      0.000000               0          0.009182                  10   \n",
       "63690      0.000307               2          0.011019                   6   \n",
       "\n",
       "       mass_match  s2v_score  label  prediction  \n",
       "66633         0.0   0.825818      1           1  \n",
       "66795         0.0   0.770255      1           1  \n",
       "66678         0.0   0.697615      1           1  \n",
       "64335         0.0   0.654275      0           0  \n",
       "66152         0.0   0.653601      0           0  \n",
       "66722         0.0   0.653601      0           0  \n",
       "66744         0.0   0.651371      0           0  \n",
       "64311         0.0   0.651371      0           0  \n",
       "66727         0.0   0.639592      0           0  \n",
       "68782         0.0   0.635570      1           0  \n",
       "62839         0.0   0.632343      0           0  \n",
       "64229         0.0   0.626987      0           0  \n",
       "62566         0.0   0.624334      0           0  \n",
       "64271         0.0   0.609262      0           0  \n",
       "68768         0.0   0.608923      0           0  \n",
       "63474         0.0   0.607861      0           0  \n",
       "63569         0.0   0.607152      0           0  \n",
       "63564         0.0   0.592299      0           0  \n",
       "64097         0.0   0.590802      0           0  \n",
       "63690         0.0   0.586521      0           0  "
      ]
     },
     "execution_count": 180,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ind = 18\n",
    "predi = all_result[3].predict(found_matches_s2v[ind])\n",
    "dfi = found_matches_s2v_labels[ind].copy()\n",
    "dfi['prediction'] = predi\n",
    "sorts2v = dfi.sort_values(by=['s2v_score'], ascending=False).iloc[:20]\n",
    "sorts2v"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 181,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'spectrum_id': 'CCMSLIB00000221190', 'source_file': 'Massbank_ESI_positive_8_1_2014_peaks.mgf', 'task': '2b0fcfed18d74148aa284e857bb00e4d', 'scan': '-1', 'ms_level': '2', 'library_membership': 'MASSBANK', 'spectrum_status': '1', 'splash': 'splash10-03dm-0896000000-0896000000', 'submit_user': 'mwang87', 'compound_name': 'E-64', 'ion_source': 'ESI', 'compound_source': 'Isolated', 'instrument': 'LC-ESI-IT', 'pi': 'Putative Massbank Match', 'data_collector': 'Massbank', 'adduct': '[M+H]+', 'precursor_mz': 358.0, 'exactmass': '0.0', 'charge': 1, 'cas_number': '66701-25-5', 'pubmed_id': 'N/A', 'smiles': 'NC(=N)NCCCCNC(=O)C(CC(C)C)NC(=O)C(O1)C(C(O)=O)1', 'inchi': 'InChI=1S/C15H27N5O5/c1-8(2)7-9(20-13(22)10-11(25-10)14(23)24)12(21)18-5-3-4-6-19-15(16)17/h8-11H,3-7H2,1-2H3,(H,18,21)(H,20,22)(H,23,24)(H4,16,17,19)', 'inchiaux': 'N/A', 'library_class': '3', 'spectrumid': 'CCMSLIB00000221190', 'ionmode': 'positive', 'create_time': '2014-08-01 16:31:38.0', 'task_id': '2b0fcfed18d74148aa284e857bb00e4d', 'user_id': 'null', 'inchikey_smiles': 'LTLYEAJONXGNFG-UHFFFAOYSA-N', 'inchikey_inchi': 'LTLYEAJONXGNFG-UHFFFAOYSA-N', 'formula_smiles': 'C15H27N5O5', 'formula_inchi': 'C15H27N5O5', 'url': 'https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00000221190', 'annotation_history': [{'Compound_Name': 'Massbank:KO008950 E-64', 'Ion_Source': 'ESI', 'Compound_Source': 'Isolated', 'Instrument': 'LC-ESI-IT', 'PI': 'Putative Massbank Match', 'Data_Collector': 'Massbank', 'Adduct': '[M+H]+', 'Scan': '-1', 'Precursor_MZ': '358.0', 'ExactMass': '0.0', 'Charge': '1', 'CAS_Number': '66701-25-5', 'Pubmed_ID': 'N/A', 'Smiles': 'NC(=N)NCCCCNC(=O)C(CC(C)C)NC(=O)C(O1)C(C(O)=O)1', 'INCHI': '1S/C15H27N5O5/c1-8(2)7-9(20-13(22)10-11(25-10)14(23)24)12(21)18-5-3-4-6-19-15(16)17/h8-11H,3-7H2,1-2H3,(H,18,21)(H,20,22)(H,23,24)(H4,16,17,19)', 'INCHI_AUX': 'N/A', 'Library_Class': '3', 'SpectrumID': 'CCMSLIB00000221190', 'Ion_Mode': 'Positive', 'create_time': '2014-08-01 16:31:38.0', 'task_id': '2b0fcfed18d74148aa284e857bb00e4d', 'user_id': 'mwang87'}], 'parent_mass': 356.9927235480092, 'inchikey': 'LTLYEAJONXGNFG-UHFFFAOYSA-N'}\n",
      "{'spectrum_id': 'CCMSLIB00000567228', 'source_file': 'output_mgf.mgf', 'task': '66fa60546df04f46bbaf8e51e61cfc2c', 'scan': '-1', 'ms_level': '2', 'library_membership': 'MONA', 'spectrum_status': '1', 'splash': 'splash10-03dl-0590000000-0590000000', 'submit_user': 'mwang87', 'compound_name': 'E-64', 'ion_source': 'LC-ESI-IT', 'compound_source': 'isolated', 'instrument': 'LC-ESI-IT', 'pi': 'MoNA', 'data_collector': 'MoNA', 'adduct': '[M+H]+', 'precursor_mz': 330.0, 'exactmass': '0.0', 'charge': 1, 'cas_number': 'N/A', 'pubmed_id': 'N/A', 'smiles': 'CC(C)CC(N=C(O)C1OC1C(=O)O)C(O)=NCCCCNC(=N)N', 'inchi': 'InChI=1S/C15H27N5O5/c1-8(2)7-9(20-13(22)10-11(25-10)14(23)24)12(21)18-5-3-4-6-19-15(16)17/h8-11H,3-7H2,1-2H3,(H,18,21)(H,20,22)(H,23,24)(H4,16,17,19)', 'inchiaux': 'N/A', 'library_class': '3', 'spectrumid': 'CCMSLIB00000567228', 'ionmode': 'positive', 'create_time': '2015-12-22 11:02:58.0', 'task_id': '66fa60546df04f46bbaf8e51e61cfc2c', 'user_id': 'null', 'inchikey_smiles': 'LTLYEAJONXGNFG-UHFFFAOYSA-N', 'inchikey_inchi': '', 'formula_smiles': 'C15H27N5O5', 'formula_inchi': '', 'url': 'https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00000567228', 'annotation_history': [{'Compound_Name': 'MoNA:2459579 E-64', 'Ion_Source': 'LC-ESI-IT', 'Compound_Source': 'isolated', 'Instrument': 'LC-ESI-IT', 'PI': 'MoNA', 'Data_Collector': 'MoNA', 'Adduct': '[M+H]+', 'Scan': '-1', 'Precursor_MZ': '330.0', 'ExactMass': '0.0', 'Charge': '1', 'CAS_Number': 'N/A', 'Pubmed_ID': 'N/A', 'Smiles': 'N/A', 'INCHI': 'InChI=1S/C15H27N5O5/c1-8(2)7-9(20-13(22)10-11(25-10)14(23)24)12(21)18-5-3-4-6-19-15(16)17/h8-11H,3-7H2,1-2H3,(H,18,21)(H,20,22)(H,23,24)(H4,16,17,19)', 'INCHI_AUX': 'N/A', 'Library_Class': '3', 'SpectrumID': 'CCMSLIB00000567228', 'Ion_Mode': '', 'create_time': '2015-12-22 11:02:58.0', 'task_id': '66fa60546df04f46bbaf8e51e61cfc2c', 'user_id': 'mwang87'}], 'parent_mass': 328.9927235480092, 'inchikey': 'LTLYEAJONXGNFG-UHFFFAOYSA-N'}\n"
     ]
    }
   ],
   "source": [
    "#check out query 18 which does not have mass matches but is a true match\n",
    "print(documents_query_s2v[18]._obj._metadata)\n",
    "print((documents_library_s2v[66633]._obj._metadata))\n",
    "##they are different..? but the same compound? how.\n",
    "smiles1 = documents_query_s2v[18]._obj._metadata.get(\"smiles\")\n",
    "smiles2 = documents_library_s2v[66633]._obj._metadata.get(\"smiles\")\n",
    "smiles3 = documents_library_s2v[66795]._obj._metadata.get(\"smiles\")\n",
    "\n",
    "from rdkit import Chem, DataStructs\n",
    "ms = [Chem.MolFromSmiles(smiles1), Chem.MolFromSmiles(smiles2), Chem.MolFromSmiles(smiles3)]\n",
    "fps = [Chem.RDKFingerprint(x) for x in ms]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Inchikeys are matching: True\n",
      "Parent masses: 356.9927235480092 328.9927235480092 339.9927235480092\n"
     ]
    },
    {
     "data": {
      "image/svg+xml": [
       "<svg baseProfile=\"full\" height=\"300px\" version=\"1.1\" viewBox=\"0 0 900 300\" width=\"900px\" xml:space=\"preserve\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:rdkit=\"http://www.rdkit.org/xml\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n",
       "<!-- END OF HEADER -->\n",
       "<rect height=\"300\" style=\"opacity:1.0;fill:#FFFFFF;stroke:none\" width=\"900\" x=\"0\" y=\"0\"> </rect>\n",
       "<rect height=\"300\" style=\"opacity:1.0;fill:#FFFFFF;stroke:none\" width=\"900\" x=\"0\" y=\"0\"> </rect>\n",
       "<rect height=\"300\" style=\"opacity:1.0;fill:#FFFFFF;stroke:none\" width=\"900\" x=\"0\" y=\"0\"> </rect>\n",
       "<rect height=\"300\" style=\"opacity:1.0;fill:#FFFFFF;stroke:none\" width=\"900\" x=\"0\" y=\"0\"> </rect>\n",
       "<path class=\"bond-0\" d=\"M 272.266,117.05 L 265.933,122.406\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-0\" d=\"M 265.933,122.406 L 259.599,127.762\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-1\" d=\"M 257.586,128.125 L 259.093,136.487\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-1\" d=\"M 259.093,136.487 L 260.601,144.849\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-1\" d=\"M 261.613,127.399 L 263.12,135.761\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-1\" d=\"M 263.12,135.761 L 264.628,144.123\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-2\" d=\"M 259.599,127.762 L 251.451,124.831\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-2\" d=\"M 251.451,124.831 L 243.303,121.901\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-3\" d=\"M 237.392,123.337 L 231.059,128.693\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-3\" d=\"M 231.059,128.693 L 224.725,134.049\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-4\" d=\"M 224.725,134.049 L 205.473,127.125\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-5\" d=\"M 205.473,127.125 L 189.851,140.336\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-6\" d=\"M 189.851,140.336 L 170.599,133.412\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-7\" d=\"M 170.599,133.412 L 164.266,138.768\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-7\" d=\"M 164.266,138.768 L 157.933,144.124\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-8\" d=\"M 152.022,145.56 L 143.874,142.63\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-8\" d=\"M 143.874,142.63 L 135.725,139.7\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-9\" d=\"M 137.739,139.337 L 136.231,130.974\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-9\" d=\"M 136.231,130.974 L 134.724,122.612\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-9\" d=\"M 133.712,140.063 L 132.204,131.7\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-9\" d=\"M 132.204,131.7 L 130.697,123.338\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-10\" d=\"M 135.725,139.7 L 120.103,152.911\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-11\" d=\"M 120.103,152.911 L 123.733,173.045\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-15\" d=\"M 120.103,152.911 L 111.955,149.98\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-15\" d=\"M 111.955,149.98 L 103.807,147.05\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-12\" d=\"M 123.733,173.045 L 142.985,179.969\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-13\" d=\"M 142.985,179.969 L 146.615,200.103\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-14\" d=\"M 142.985,179.969 L 158.607,166.758\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-16\" d=\"M 100.237,142.577 L 98.7292,134.215\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-16\" d=\"M 98.7292,134.215 L 97.2216,125.852\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-17\" d=\"M 98.5427,127.415 L 104.761,122.156\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-17\" d=\"M 104.761,122.156 L 110.98,116.897\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-17\" d=\"M 95.9005,124.29 L 102.119,119.031\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-17\" d=\"M 102.119,119.031 L 108.338,113.773\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-18\" d=\"M 97.2216,125.852 L 77.9697,118.929\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-19\" d=\"M 77.9697,118.929 L 72.806,112.823\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-19\" d=\"M 72.806,112.823 L 67.6424,106.717\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-20\" d=\"M 77.9697,118.929 L 57.8352,122.559\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-24\" d=\"M 63.5325,106.717 L 60.6839,114.638\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-24\" d=\"M 60.6839,114.638 L 57.8352,122.559\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-21\" d=\"M 57.8352,122.559 L 42.2132,135.769\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-22\" d=\"M 42.2132,135.769 L 34.0649,132.839\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-22\" d=\"M 34.0649,132.839 L 25.9167,129.909\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-23\" d=\"M 40.1997,136.132 L 41.7073,144.495\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-23\" d=\"M 41.7073,144.495 L 43.2149,152.857\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-23\" d=\"M 44.2266,135.407 L 45.7342,143.769\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-23\" d=\"M 45.7342,143.769 L 47.2418,152.131\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"start\" x=\"273.016\" y=\"115.676\"><tspan>NH</tspan><tspan style=\"baseline-shift:sub;font-size:4.5px;\">2</tspan><tspan/></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"start\" x=\"260.956\" y=\"148.919\"><tspan>NH</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"middle\" x=\"240.347\" y=\"121.861\"><tspan>N</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"start\" x=\"238.074\" y=\"115.041\"><tspan>H</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"middle\" x=\"154.977\" y=\"147.646\"><tspan>N</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"start\" x=\"152.704\" y=\"154.466\"><tspan>H</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"start\" x=\"129.593\" y=\"120.588\"><tspan>O</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"end\" x=\"103.125\" y=\"147.01\"><tspan>HN</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"start\" x=\"110.341\" y=\"113.664\"><tspan>O</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"middle\" x=\"64.7588\" y=\"104.33\"><tspan>O</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"end\" x=\"25.464\" y=\"129.869\"><tspan>HO</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"start\" x=\"43.3404\" y=\"156.927\"><tspan>O</tspan></text>\n",
       "<path class=\"bond-0\" d=\"M 446.615,200.103 L 442.985,179.969\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-1\" d=\"M 442.985,179.969 L 458.607,166.758\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-2\" d=\"M 442.985,179.969 L 423.733,173.045\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-3\" d=\"M 423.733,173.045 L 420.103,152.911\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-4\" d=\"M 420.103,152.911 L 411.955,149.98\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-4\" d=\"M 411.955,149.98 L 403.807,147.05\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-13\" d=\"M 420.103,152.911 L 435.725,139.7\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-5\" d=\"M 400.237,142.577 L 398.729,134.215\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-5\" d=\"M 398.729,134.215 L 397.222,125.852\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-5\" d=\"M 403.811,139.342 L 402.756,133.489\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-5\" d=\"M 402.756,133.489 L 401.701,127.635\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-6\" d=\"M 397.222,125.852 L 403.44,120.594\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-6\" d=\"M 403.44,120.594 L 409.659,115.335\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-7\" d=\"M 397.222,125.852 L 377.97,118.929\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-8\" d=\"M 377.97,118.929 L 372.806,112.823\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-8\" d=\"M 372.806,112.823 L 367.642,106.717\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-24\" d=\"M 377.97,118.929 L 357.835,122.559\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-9\" d=\"M 363.533,106.717 L 360.684,114.638\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-9\" d=\"M 360.684,114.638 L 357.835,122.559\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-10\" d=\"M 357.835,122.559 L 342.213,135.769\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-11\" d=\"M 340.2,136.132 L 341.707,144.495\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-11\" d=\"M 341.707,144.495 L 343.215,152.857\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-11\" d=\"M 344.227,135.407 L 345.734,143.769\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-11\" d=\"M 345.734,143.769 L 347.242,152.131\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-12\" d=\"M 342.213,135.769 L 334.065,132.839\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-12\" d=\"M 334.065,132.839 L 325.917,129.909\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-14\" d=\"M 435.725,139.7 L 434.218,131.337\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-14\" d=\"M 434.218,131.337 L 432.71,122.975\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-15\" d=\"M 435.725,139.7 L 443.874,142.63\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-15\" d=\"M 443.874,142.63 L 452.022,145.56\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-15\" d=\"M 436.785,144.429 L 442.489,146.48\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-15\" d=\"M 442.489,146.48 L 448.193,148.532\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-16\" d=\"M 457.933,144.124 L 464.266,138.768\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-16\" d=\"M 464.266,138.768 L 470.599,133.412\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-17\" d=\"M 470.599,133.412 L 489.851,140.336\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-18\" d=\"M 489.851,140.336 L 505.473,127.125\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-19\" d=\"M 505.473,127.125 L 524.725,134.049\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-20\" d=\"M 524.725,134.049 L 531.059,128.693\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-20\" d=\"M 531.059,128.693 L 537.392,123.337\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-21\" d=\"M 543.303,121.901 L 551.451,124.831\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-21\" d=\"M 551.451,124.831 L 559.599,127.762\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-22\" d=\"M 557.586,128.125 L 559.093,136.487\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-22\" d=\"M 559.093,136.487 L 560.601,144.849\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-22\" d=\"M 561.613,127.399 L 563.12,135.761\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-22\" d=\"M 563.12,135.761 L 564.628,144.123\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-23\" d=\"M 559.599,127.762 L 565.933,122.406\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-23\" d=\"M 565.933,122.406 L 572.266,117.05\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"end\" x=\"403.125\" y=\"147.01\"><tspan>N</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"start\" x=\"410.341\" y=\"113.664\"><tspan>OH</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"middle\" x=\"364.759\" y=\"104.33\"><tspan>O</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"start\" x=\"343.34\" y=\"156.927\"><tspan>O</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"end\" x=\"325.464\" y=\"129.869\"><tspan>HO</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"start\" x=\"429.593\" y=\"120.588\"><tspan>OH</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"middle\" x=\"454.977\" y=\"147.646\"><tspan>N</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"middle\" x=\"540.347\" y=\"121.861\"><tspan>N</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"start\" x=\"538.074\" y=\"115.041\"><tspan>H</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"start\" x=\"560.956\" y=\"148.919\"><tspan>NH</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"start\" x=\"573.016\" y=\"115.676\"><tspan>NH</tspan><tspan style=\"baseline-shift:sub;font-size:4.5px;\">2</tspan><tspan/></text>\n",
       "<path class=\"bond-0\" d=\"M 746.615,200.103 L 742.985,179.969\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-1\" d=\"M 742.985,179.969 L 758.607,166.758\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-2\" d=\"M 742.985,179.969 L 723.733,173.045\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-3\" d=\"M 723.733,173.045 L 720.103,152.911\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-4\" d=\"M 720.103,152.911 L 711.955,149.98\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-4\" d=\"M 711.955,149.98 L 703.807,147.05\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-13\" d=\"M 720.103,152.911 L 735.725,139.7\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-5\" d=\"M 700.237,142.577 L 698.729,134.215\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-5\" d=\"M 698.729,134.215 L 697.222,125.852\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-5\" d=\"M 703.811,139.342 L 702.756,133.489\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-5\" d=\"M 702.756,133.489 L 701.701,127.635\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-6\" d=\"M 697.222,125.852 L 703.44,120.594\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-6\" d=\"M 703.44,120.594 L 709.659,115.335\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-7\" d=\"M 697.222,125.852 L 677.97,118.929\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-8\" d=\"M 677.97,118.929 L 672.806,112.823\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-8\" d=\"M 672.806,112.823 L 667.642,106.717\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-24\" d=\"M 677.97,118.929 L 657.835,122.559\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-9\" d=\"M 663.533,106.717 L 660.684,114.638\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-9\" d=\"M 660.684,114.638 L 657.835,122.559\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-10\" d=\"M 657.835,122.559 L 642.213,135.769\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-11\" d=\"M 640.2,136.132 L 641.707,144.495\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-11\" d=\"M 641.707,144.495 L 643.215,152.857\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-11\" d=\"M 644.227,135.407 L 645.734,143.769\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-11\" d=\"M 645.734,143.769 L 647.242,152.131\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-12\" d=\"M 642.213,135.769 L 634.065,132.839\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-12\" d=\"M 634.065,132.839 L 625.917,129.909\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-14\" d=\"M 735.725,139.7 L 734.218,131.337\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-14\" d=\"M 734.218,131.337 L 732.71,122.975\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-15\" d=\"M 735.725,139.7 L 743.874,142.63\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-15\" d=\"M 743.874,142.63 L 752.022,145.56\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-15\" d=\"M 736.785,144.429 L 742.489,146.48\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-15\" d=\"M 742.489,146.48 L 748.193,148.532\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-16\" d=\"M 757.933,144.124 L 764.266,138.768\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-16\" d=\"M 764.266,138.768 L 770.599,133.412\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-17\" d=\"M 770.599,133.412 L 789.851,140.336\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-18\" d=\"M 789.851,140.336 L 805.473,127.125\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-19\" d=\"M 805.473,127.125 L 824.725,134.049\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-20\" d=\"M 824.725,134.049 L 831.059,128.693\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-20\" d=\"M 831.059,128.693 L 837.392,123.337\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-21\" d=\"M 843.303,121.901 L 851.451,124.831\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-21\" d=\"M 851.451,124.831 L 859.599,127.762\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-22\" d=\"M 857.586,128.125 L 859.093,136.487\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-22\" d=\"M 859.093,136.487 L 860.601,144.849\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-22\" d=\"M 861.613,127.399 L 863.12,135.761\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-22\" d=\"M 863.12,135.761 L 864.628,144.123\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-23\" d=\"M 859.599,127.762 L 865.933,122.406\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path class=\"bond-23\" d=\"M 865.933,122.406 L 872.266,117.05\" style=\"fill:none;fill-rule:evenodd;stroke:#0000FF;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"end\" x=\"703.125\" y=\"147.01\"><tspan>N</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"start\" x=\"710.341\" y=\"113.664\"><tspan>OH</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"middle\" x=\"664.759\" y=\"104.33\"><tspan>O</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"start\" x=\"643.34\" y=\"156.927\"><tspan>O</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"end\" x=\"625.464\" y=\"129.869\"><tspan>HO</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#FF0000\" text-anchor=\"start\" x=\"729.593\" y=\"120.588\"><tspan>OH</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"middle\" x=\"754.977\" y=\"147.646\"><tspan>N</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"middle\" x=\"840.347\" y=\"121.861\"><tspan>N</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"start\" x=\"838.074\" y=\"115.041\"><tspan>H</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"start\" x=\"860.956\" y=\"148.919\"><tspan>NH</tspan></text>\n",
       "<text dominant-baseline=\"central\" style=\"font-size:6px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;fill:#0000FF\" text-anchor=\"start\" x=\"873.016\" y=\"115.676\"><tspan>NH</tspan><tspan style=\"baseline-shift:sub;font-size:4.5px;\">2</tspan><tspan/></text>\n",
       "</svg>"
      ],
      "text/plain": [
       "<IPython.core.display.SVG object>"
      ]
     },
     "execution_count": 182,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"Inchikeys are matching:\",\n",
    "      documents_query_s2v[18]._obj._metadata.get(\"inchikey\") == documents_library_s2v[66633]._obj._metadata.get(\"inchikey\"))\n",
    "print(\"Parent masses:\",\n",
    "      documents_query_s2v[18]._obj._metadata.get(\"parent_mass\"),\n",
    "      documents_library_s2v[66633]._obj._metadata.get(\"parent_mass\"),\n",
    "      documents_library_s2v[66795]._obj._metadata.get(\"parent_mass\"))\n",
    "img = Chem.Draw.MolsToGridImage(ms, molsPerRow=3, subImgSize=(300, 300), maxMols=len(ms),\n",
    "                           legends=None, useSVG=True)\n",
    "img"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}