forked from UMCUGenetics/pymc4c
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathanalysis.py
707 lines (603 loc) · 29.8 KB
/
analysis.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
import numpy as np
from utilities import showprogress
MIN_N_POS = 100
def compute_mc_associations(frg_inf, pos_crd, bin_bnd, n_perm=1000, pos_ids=None, verbose=True):
from utilities import hasOL, flatten
# initialization
n_bin = bin_bnd.shape[0]
# re-index circles
frg_inf[:, 0] = np.unique(frg_inf[:, 0], return_inverse=True)[1] + 1
n_read = np.max(frg_inf[:, 0])
# convert fragments to bin-coverage
cfb_lst = [list() for i in range(n_read + 1)]
n_frg = frg_inf.shape[0]
for fi in range(n_frg):
bin_idx = np.where(hasOL(frg_inf[fi, 2:4], bin_bnd))[0]
cfb_lst[frg_inf[fi, 0]].append(list(bin_idx))
# select positive/negative circles
if pos_ids is not None:
assert len(pos_crd) == 0
is_pos = np.isin(frg_inf[:, 0], pos_ids)
else:
is_pos = np.where(hasOL(pos_crd, frg_inf[:, 1:4]))[0]
frg_pos = frg_inf[ np.isin(frg_inf[:, 0], frg_inf[is_pos, 0]), :]
frg_neg = frg_inf[~np.isin(frg_inf[:, 0], frg_inf[is_pos, 0]), :]
cfb_pos = [cfb_lst[i] for i in np.unique(frg_pos[:, 0])]
cfb_neg = [cfb_lst[i] for i in np.unique(frg_neg[:, 0])]
n_pos = len(cfb_pos)
n_neg = len(cfb_neg)
# make positive profile
prf_pos = np.zeros(n_bin)
for pi in range(n_pos):
bin_lst = flatten(cfb_pos[pi])
prf_pos[bin_lst] += 1
# make background profile from negative set
prf_rnd = np.zeros([n_perm, n_bin])
neg_lst = range(n_neg)
for ei in np.arange(n_perm):
if verbose and (((ei + 1) % 200) == 0):
print '\t{:d} randomized profiles are computed.'.format(ei + 1)
np.random.shuffle(neg_lst)
for rd_idx in neg_lst[:n_pos]:
f2b_rnd = cfb_neg[rd_idx]
np.random.shuffle(f2b_rnd)
prf_rnd[ei, flatten(f2b_rnd[1:])] += 1 # making sure one element is randomly removed everytime
return prf_pos, prf_rnd, frg_pos, frg_neg
def perform_mc_analysis(configs, min_n_frg=2):
import platform
if platform.system() == 'Linux':
import matplotlib
matplotlib.use('Agg')
from matplotlib import pyplot as plt, patches
from matplotlib.colors import LinearSegmentedColormap
from utilities import load_mc4c, load_annotation, hasOL
# initialization
if configs['output_file'] is None:
configs['output_file'] = configs['output_dir'] + '/analysis_mcTest_' + configs['run_id'] + '.pdf'
edge_lst = np.linspace(configs['roi_start'], configs['roi_end'], num=201, dtype=np.int64).reshape(-1, 1)
bin_bnd = np.hstack([edge_lst[:-1], edge_lst[1:] - 1])
n_bin = bin_bnd.shape[0]
n_epoch = 1000
x_lim = [configs['roi_start'], configs['roi_end']]
# load MC-HC data
frg_dp = load_mc4c(configs, unique_only=True, valid_only=True, min_mq=20, reindex_reads=False)
frg_np = frg_dp[['ReadID', 'Chr', 'ExtStart', 'ExtEnd']].values
del frg_dp
# select within roi fragments
vp_crd = [configs['vp_cnum'], configs['vp_start'], configs['vp_end']]
roi_crd = [configs['vp_cnum'], configs['roi_start'], configs['roi_end']]
is_vp = hasOL(vp_crd, frg_np[:, 1:4])
is_roi = hasOL(roi_crd, frg_np[:, 1:4])
frg_roi = frg_np[~is_vp & is_roi, :]
del frg_np
# filter small circles
cir_size = np.bincount(frg_roi[:, 0])[frg_roi[:, 0]]
frg_roi = frg_roi[cir_size >= min_n_frg, :]
n_read = len(np.unique(frg_roi[:, 0]))
# re-index circles
frg_roi[:, 0] = np.unique(frg_roi[:, 0], return_inverse=True)[1]
# convert reads to bin coverage
cvg_lst = [list() for i in range(n_read)]
for fi in range(frg_roi.shape[0]):
bin_idx = np.where(hasOL(frg_roi[fi, 2:4], bin_bnd))[0]
cvg_lst[frg_roi[fi, 0]].extend(bin_idx)
cvg_lst = [np.unique(cvg_lst[i]) for i in range(n_read)]
# looping over bins
print 'Performing the MC analysis using {:d} reads ...'.format(n_read)
mat_freq = np.full([n_bin, n_bin], fill_value=np.nan)
mat_zscr = np.full([n_bin, n_bin], fill_value=np.nan)
for bi in range(n_bin):
if bi % (n_bin / 10) == 0:
print '{:0.0f}%,'.format(bi * 100.0 / n_bin),
is_pos = hasOL(bin_bnd[bi, :], frg_roi[:, 2:4])
frg_pos = frg_roi[np.isin(frg_roi[:, 0], frg_roi[is_pos, 0]), :]
frg_neg = frg_roi[~np.isin(frg_roi[:, 0], frg_pos[:, 0]), :]
ids_pos = np.unique(frg_pos[:, 0])
ids_neg = np.unique(frg_neg[:, 0])
n_pos = len(ids_pos)
n_neg = len(ids_neg)
assert n_pos <= n_neg
if n_pos < 100:
continue
# calculate the background
rnd_freq = np.zeros([n_epoch, n_bin])
for ei in np.arange(n_epoch):
rnd_lst = np.random.choice(ids_neg, n_pos, replace=False)
for rd_idx in rnd_lst:
bin_cvg = cvg_lst[rd_idx]
n_cvg = len(bin_cvg)
rnd_freq[ei, bin_cvg] += 1
rnd_freq[ei, bin_cvg[np.random.randint(n_cvg)]] -= 1
# calculate observed
for bj in range(bi+1, n_bin):
is_cov = hasOL(bin_bnd[bj, :], frg_pos[:, 2:4])
mat_freq[bi, bj] = len(np.unique(frg_pos[is_cov, 0]))
zscr_avg = np.mean(rnd_freq[:, bj])
zscr_std = np.std(rnd_freq[:, bj])
if zscr_std == 0:
continue
mat_zscr[bi, bj] = (mat_freq[bi, bj] - zscr_avg) / zscr_std
mat_zscr[bj, bi] = mat_zscr[bi, bj]
# set vp bins to nan
is_vp = hasOL([configs['vp_start'], configs['vp_end']], bin_bnd)
mat_zscr[is_vp, :] = np.nan
mat_zscr[:, is_vp] = np.nan
vp_bnd = [bin_bnd[is_vp, 0][0], bin_bnd[is_vp, 1][-1]]
# plotting
plt.figure(figsize=(17, 9))
clr_lst = ['#ff1a1a', '#ff8a8a', '#ffffff', '#ffffff', '#ffffff', '#8ab5ff', '#3900f5']
clr_map = LinearSegmentedColormap.from_list('test', clr_lst, N=10)
clr_map.set_bad('gray', 0.05)
plt.imshow(mat_zscr, extent=x_lim + x_lim, cmap=clr_map, origin='bottom', interpolation='nearest')
plt.gca().add_patch(patches.Rectangle([vp_bnd[0], x_lim[0]], vp_bnd[1] - vp_bnd[0], x_lim[1] - x_lim[0],
linewidth=0, edgecolor='None', facecolor='orange'))
plt.gca().add_patch(patches.Rectangle([x_lim[0], vp_bnd[0]], x_lim[1] - x_lim[0], vp_bnd[1] - vp_bnd[0],
linewidth=0, edgecolor='None', facecolor='orange'))
cbar_h = plt.colorbar()
cbar_h.ax.tick_params(labelsize=14)
plt.clim(-6, 6)
# add annotations
ant_pd = load_annotation(configs['genome_build'], roi_crd=[configs['vp_cnum'], configs['roi_start'], configs['roi_end']])
for ai in range(ant_pd.shape[0]):
ant_pos = ant_pd.loc[ai, 'ant_pos']
plt.text(ant_pos, x_lim[1], ant_pd.loc[ai, 'ant_name'],
horizontalalignment='left', verticalalignment='bottom', rotation=60)
plt.text(x_lim[1], ant_pos, ' ' + ant_pd.loc[ai, 'ant_name'],
horizontalalignment='left', verticalalignment='center')
plt.plot([ant_pos, ant_pos], x_lim, ':', color='#bfbfbf', linewidth=1, alpha=0.4)
plt.plot(x_lim, [ant_pos, ant_pos], ':', color='#bfbfbf', linewidth=1, alpha=0.4)
# final adjustments
plt.xlim(x_lim)
plt.ylim(x_lim)
x_ticks = np.linspace(configs['roi_start'], configs['roi_end'], 7, dtype=np.int64)
x_tick_label = ['{:0.2f}m'.format(x / 1e6) for x in x_ticks]
plt.xticks(x_ticks, x_tick_label, rotation=0, horizontalalignment='center')
plt.yticks(x_ticks, x_tick_label, rotation=0)
plt.title('Multicontact matrix, {:s}\n'.format(configs['run_id']) +
'#read (#roiFrg>{:d}, ex. vp)={:,d}\n\n\n'.format(min_n_frg - 1, n_read)
)
plt.savefig(configs['output_file'], bbox_inches='tight')
def perform_vpsoi_analysis(configs, soi_name, min_n_frg=2, n_perm=1000):
import platform
import matplotlib
if platform.system() == 'Linux':
matplotlib.use('Agg')
from matplotlib import pyplot as plt, patches
from matplotlib.colors import LinearSegmentedColormap
from utilities import load_mc4c, load_annotation, hasOL, flatten
# initialization
if configs['output_file'] is None:
configs['output_file'] = configs['output_dir'] + '/analysis_atVP-SOI_{:s}_{:s}.pdf'.format(configs['run_id'], soi_name)
edge_lst = np.linspace(configs['roi_start'], configs['roi_end'], num=201, dtype=np.int64).reshape(-1, 1)
bin_bnd = np.hstack([edge_lst[:-1], edge_lst[1:] - 1])
bin_cen = np.mean(bin_bnd, axis=1, dtype=np.int64)
bin_w = bin_bnd[0, 1] - bin_bnd[0, 0]
x_lim = [configs['roi_start'], configs['roi_end']]
y_lim = [0, 10]
# load MC-HC data
frg_dp = load_mc4c(configs, unique_only=True, valid_only=True, min_mq=20, reindex_reads=True, verbose=True)
frg_np = frg_dp[['ReadID', 'Chr', 'ExtStart', 'ExtEnd']].values
del frg_dp
# select within roi fragments
vp_crd = [configs['vp_cnum'], configs['vp_start'], configs['vp_end']]
roi_crd = [configs['vp_cnum'], configs['roi_start'], configs['roi_end']]
is_vp = hasOL(vp_crd, frg_np[:, 1:4])
is_roi = hasOL(roi_crd, frg_np[:, 1:4])
frg_roi = frg_np[~is_vp & is_roi, :]
del frg_np
# filter small circles (>1 roi-frg, ex.)
cir_size = np.bincount(frg_roi[:, 0])[frg_roi[:, 0]]
frg_inf = frg_roi[cir_size >= min_n_frg, :]
frg_inf[:, 0] = np.unique(frg_inf[:, 0], return_inverse=True)[1] + 1
n_read = len(np.unique(frg_inf[:, 0]))
# convert fragments to bin-coverage
cfb_lst = [list() for i in range(n_read + 1)]
n_frg = frg_inf.shape[0]
for fi in range(n_frg):
bin_idx = np.where(hasOL(frg_inf[fi, 2:4], bin_bnd))[0]
cfb_lst[frg_inf[fi, 0]].append(bin_idx.tolist())
# filter circles for (>1 bin cvg)
valid_lst = []
for rd_nid in range(1, n_read + 1):
fb_lst = cfb_lst[rd_nid]
bin_cvg = np.unique(flatten(fb_lst))
if len(bin_cvg) > 1:
valid_lst.append(rd_nid)
frg_inf = frg_inf[np.isin(frg_inf[:, 0], valid_lst), :]
frg_inf[:, 0] = np.unique(frg_inf[:, 0], return_inverse=True)[1] + 1
n_read = np.max(frg_inf[:, 0])
# get soi info
ant_pd = load_annotation(configs['genome_build'], roi_crd=roi_crd).reset_index(drop=True)
n_ant = ant_pd.shape[0]
is_in = np.where(np.isin(ant_pd['ant_name'], soi_name))[0]
assert len(is_in) == 1
soi_pd = ant_pd.loc[is_in[0], :]
soi_crd = [soi_pd['ant_cnum'], soi_pd['ant_pos'] - int(bin_w * 1.5), soi_pd['ant_pos'] + int(bin_w * 1.5)]
if hasOL(soi_crd, vp_crd)[0]:
'[w] Selected SOI coordinate overlaps with the view point. Ignoring the analysis'
return
# compute positive profile and backgrounds
print 'Computing expected profile for bins:'
prf_frq, prf_rnd, frg_pos, frg_neg = compute_mc_associations(frg_inf, soi_crd, bin_bnd, n_perm=n_perm)
n_pos = len(np.unique(frg_pos[:, 0]))
prf_obs = prf_frq * 100.0 / n_pos
print '{:,d} reads are found to cover '.format(n_pos) + \
'{:s} area ({:s}:{:d}-{:d})'.format(soi_pd['ant_name'], soi_pd['ant_chr'], soi_crd[1], soi_crd[2])
# check enough #pos
if n_pos < MIN_N_POS:
print '[w] #reads in the positive set is insufficient (n={:d}, required >{:d})'.format(n_pos, MIN_N_POS)
print 'Analysis is ignored ...'
return
# compute scores
nrm_rnd = prf_rnd * 100.0 / n_pos
prf_exp = np.mean(nrm_rnd, axis=0)
prf_std = np.std(nrm_rnd, axis=0, ddof=0)
np.seterr(all='ignore')
bin_scr = np.divide(prf_obs - prf_exp, prf_std)
np.seterr(all=None)
# set vp bins to nan
vp_bnd = [configs['vp_start'], configs['vp_end']]
is_vp = hasOL(vp_bnd, bin_bnd)
bin_scr[is_vp] = np.nan
# compute score for annotations
print 'Computing expected profile for annotations:'
ant_pos = ant_pd['ant_pos'].values.reshape(-1, 1)
ant_bnd = np.hstack([ant_pos - int(bin_w * 1.5), ant_pos + int(bin_w * 1.5)])
ant_obs, soi_rnd = compute_mc_associations(frg_inf, soi_crd, ant_bnd, n_perm=n_perm)[:2]
ant_exp = np.mean(soi_rnd, axis=0)
ant_std = np.std(soi_rnd, axis=0, ddof=0)
np.seterr(all='ignore')
ant_scr = np.divide(ant_obs - ant_exp, ant_std)
np.seterr(all=None)
# set vp score to nan
is_vp = hasOL(vp_bnd, ant_bnd)
is_soi = hasOL(soi_crd[1:3], ant_bnd)
ant_scr[is_vp | is_soi] = np.nan
# plotting
fig = plt.figure(figsize=(15, 3))
ax_prf = plt.subplot2grid((20, 40), (0, 0), rowspan=19, colspan=39)
ax_cmp = plt.subplot2grid((20, 40), (0, 39), rowspan=10, colspan=1)
ax_scr = plt.subplot2grid((20, 40), (19, 0), rowspan=1, colspan=39)
# set up colorbar
c_lim = [-6, 6]
clr_lst = ['#ff1a1a', '#ff7575', '#ffcccc', '#ffffff', '#ffffff', '#ffffff', '#ccdfff', '#3d84ff', '#3900f5']
clr_map = LinearSegmentedColormap.from_list('test', clr_lst, N=9)
clr_map.set_bad('gray', 0.05)
norm = matplotlib.colors.Normalize(vmin=c_lim[0], vmax=c_lim[1])
cbar_h = matplotlib.colorbar.ColorbarBase(ax_cmp, cmap=clr_map, norm=norm)
# cbar_h.ax.tick_params(labelsize=12)
cbar_h.ax.set_ylabel('z-score', rotation=90)
# profile plot
ax_prf.plot(bin_cen, prf_obs, color='#5757ff', linewidth=1)
ax_prf.plot(bin_cen, prf_exp, color='#cccccc', linewidth=1)
ax_prf.fill_between(bin_cen, prf_exp - prf_std, prf_exp + prf_std, color='#ebebeb', linewidth=0.2)
ax_prf.add_patch(patches.Rectangle([vp_bnd[0], y_lim[0]], vp_bnd[1] - vp_bnd[0], y_lim[1] - y_lim[0],
edgecolor='None', facecolor='orange', zorder=100))
ax_prf.add_patch(patches.Rectangle([soi_crd[1], y_lim[0]], soi_crd[2] - soi_crd[1], y_lim[1] - y_lim[0],
edgecolor='None', facecolor='green', zorder=100))
ax_prf.set_xlim(x_lim)
ax_prf.set_ylim(y_lim)
ax_prf.set_xticks([])
# add score plot
ax_scr.imshow(bin_scr.reshape(1, -1), extent=x_lim + [-500, 500], cmap=clr_map,
vmin=c_lim[0], vmax=c_lim[1], interpolation='nearest')
ax_scr.set_xlim(x_lim)
ax_scr.set_yticks([])
# add annotations
for ai in range(n_ant):
ax_prf.text(ant_pos[ai], y_lim[1], ant_pd.loc[ai, 'ant_name'],
horizontalalignment='center', verticalalignment='bottom', rotation=60)
ax_prf.plot(ant_pos[[ai, ai]], y_lim, ':', color='#bfbfbf', linewidth=1, alpha=0.4)
if not np.isnan(ant_scr[ai]):
ax_prf.add_patch(patches.Rectangle([ant_bnd[ai, 0], y_lim[1]-0.15], ant_bnd[ai, 1] - ant_bnd[ai, 0], 0.15,
edgecolor='None', facecolor=clr_map(ant_scr[ai]), zorder=10))
ax_prf.text(ant_pos[ai], y_lim[1] - 0.2, '{:+0.1f}'.format(ant_scr[ai]),
horizontalalignment='center', verticalalignment='top', fontweight='bold', fontsize=6)
# final adjustments
x_ticks = np.linspace(configs['roi_start'], configs['roi_end'], 7, dtype=np.int64)
y_ticks = ax_prf.get_yticks()
x_tick_label = ['{:0.2f}m'.format(x / 1e6) for x in x_ticks]
y_tick_label = ['{:0.0f}%'.format(y) for y in y_ticks]
ax_scr.set_xticks(x_ticks)
ax_scr.set_xticklabels(x_tick_label)
ax_prf.set_yticklabels(y_tick_label)
ax_prf.set_ylabel('Percentage of reads')
ax_prf.set_title('VP-SOI from {:s}, using as SOI {:s}\n'.format(configs['run_id'], soi_name) +
'#read (#roiFrg>{:d}, ex. vp)={:,d}, '.format(min_n_frg - 1, n_read) +
'#pos = {:d}\n#perm={:d}\n\n\n'.format(n_pos, n_perm)
)
plt.savefig(configs['output_file'], bbox_inches='tight')
def perform_soisoi_analysis(config_lst, min_n_frg=2, n_perm=1000):
import platform
import matplotlib
if platform.system() == 'Linux':
matplotlib.use('Agg')
from matplotlib import pyplot as plt
from matplotlib.colors import LinearSegmentedColormap
from utilities import load_mc4c, load_annotation, hasOL, flatten
# initialization
run_id = ','.join([config['run_id'] for config in config_lst])
if config_lst[0]['output_file'] is None:
config_lst[0]['output_file'] = config_lst[0]['output_dir'] + '/analysis_atSOI-SOI_{:s}.pdf'.format(run_id)
edge_lst = np.linspace(config_lst[0]['roi_start'], config_lst[0]['roi_end'], num=201, dtype=np.int64).reshape(-1, 1)
bin_bnd = np.hstack([edge_lst[:-1], edge_lst[1:] - 1])
bin_w = bin_bnd[0, 1] - bin_bnd[0, 0]
del edge_lst
# load MC-HC data
frg_dp = load_mc4c(config_lst, unique_only=True, valid_only=True, min_mq=20, reindex_reads=True, verbose=True)
frg_np = frg_dp[['ReadID', 'Chr', 'ExtStart', 'ExtEnd']].values
del frg_dp
# select within roi fragments
vp_crd = [config_lst[0]['vp_cnum'], config_lst[0]['vp_start'], config_lst[0]['vp_end']]
roi_crd = [config_lst[0]['vp_cnum'], config_lst[0]['roi_start'], config_lst[0]['roi_end']]
is_vp = hasOL(vp_crd, frg_np[:, 1:4])
is_roi = hasOL(roi_crd, frg_np[:, 1:4])
frg_roi = frg_np[~is_vp & is_roi, :]
del frg_np
# filter small read (>1 roi-frg, ex.)
cir_size = np.bincount(frg_roi[:, 0])[frg_roi[:, 0]]
frg_inf = frg_roi[cir_size >= min_n_frg, :]
frg_inf[:, 0] = np.unique(frg_inf[:, 0], return_inverse=True)[1] + 1
n_read = len(np.unique(frg_inf[:, 0]))
# convert fragments to bin-coverage
cfb_lst = [list() for i in range(n_read + 1)]
n_frg = frg_inf.shape[0]
for fi in range(n_frg):
bin_idx = np.where(hasOL(frg_inf[fi, 2:4], bin_bnd))[0]
cfb_lst[frg_inf[fi, 0]].append(bin_idx.tolist())
# filter reads for (>1 bin cvg)
valid_lst = []
for rd_nid in range(1, n_read + 1):
fb_lst = cfb_lst[rd_nid]
bin_cvg = np.unique(flatten(fb_lst))
if len(bin_cvg) > 1:
valid_lst.append(rd_nid)
frg_inf = frg_inf[np.isin(frg_inf[:, 0], valid_lst), :]
# Downsample and re-index
# rnd_rid = np.random.choice(np.unique(frg_inf[:, 0]), 8618, replace=False) ### random selection
# frg_inf = frg_inf[np.isin(frg_inf[:, 0], rnd_rid), :]
frg_inf[:, 0] = np.unique(frg_inf[:, 0], return_inverse=True)[1] + 1
n_read = np.max(frg_inf[:, 0])
# loop over each SOI
ant_pd = load_annotation(config_lst[0]['genome_build'], roi_crd=roi_crd).reset_index(drop=True)
n_ant = ant_pd.shape[0]
ant_name_lst = ant_pd['ant_name'].values
ant_scr = np.full(shape=[n_ant, n_ant], fill_value=np.nan)
n_pos = np.zeros(n_ant, dtype=np.int)
x_tick_lbl = []
for ai in range(n_ant):
soi_pd = ant_pd.loc[ai, :]
soi_crd = [soi_pd['ant_cnum'], soi_pd['ant_pos'] - int(bin_w * 1.5), soi_pd['ant_pos'] + int(bin_w * 1.5)]
if hasOL(vp_crd[1:], soi_crd[1:]):
x_tick_lbl.append(ant_name_lst[ai])
continue
# compute score for annotations
print 'Computing expected profile for {:s}:'.format(soi_pd['ant_name'])
ant_pos = ant_pd['ant_pos'].values.reshape(-1, 1)
ant_bnd = np.hstack([ant_pos - int(bin_w * 1.5), ant_pos + int(bin_w * 1.5)])
ant_obs, soi_rnd, frg_pos = compute_mc_associations(frg_inf, soi_crd, ant_bnd, n_perm=n_perm)[:3]
n_pos[ai] = len(np.unique(frg_pos[:, 0]))
x_tick_lbl.append('{:s}\n#{:,d}'.format(ant_name_lst[ai], n_pos[ai]))
del frg_pos
# check number of positive reads
if n_pos[ai] <= MIN_N_POS:
print '[w] #reads (n={:d}) in the positive set is insufficient '.format(n_pos[ai]) + \
'(required >{:d}). This analysis is ignored ...'.format(MIN_N_POS)
continue
# calculate expected profile
ant_exp = np.mean(soi_rnd, axis=0)
ant_std = np.std(soi_rnd, axis=0, ddof=0)
np.seterr(all='ignore')
ant_scr[:, ai] = np.divide(ant_obs - ant_exp, ant_std)
np.seterr(all=None)
# set vp score to nan
is_vp = hasOL(vp_crd[1:], ant_bnd)
is_soi = hasOL(soi_crd[1:3], ant_bnd)
ant_scr[is_vp | is_soi, ai] = np.nan
# plotting
plt.figure(figsize=(8, 7))
ax_scr = plt.subplot2grid((40, 40), (0, 0), rowspan=39, colspan=39)
ax_cmp = plt.subplot2grid((40, 40), (0, 39), rowspan=20, colspan=1)
# set up colorbar
c_lim = [-6, 6]
clr_lst = ['#ff1a1a', '#ff7575', '#ffcccc', '#ffffff', '#ffffff', '#ffffff', '#ccdfff', '#3d84ff', '#3900f5']
clr_map = LinearSegmentedColormap.from_list('test', clr_lst, N=9)
clr_map.set_bad('gray', 0.2)
norm = matplotlib.colors.Normalize(vmin=c_lim[0], vmax=c_lim[1])
cbar_h = matplotlib.colorbar.ColorbarBase(ax_cmp, cmap=clr_map, norm=norm)
# cbar_h.ax.tick_params(labelsize=12)
cbar_h.ax.set_ylabel('z-score', rotation=90)
cbar_edge = np.round(cbar_h.cmap(norm(c_lim)), decimals=2)
# add score scatter matrix
x_lim = [0, n_ant]
img_h = ax_scr.imshow(ant_scr, extent=x_lim + x_lim, cmap=clr_map,
vmin=c_lim[0], vmax=c_lim[1], interpolation='nearest', origin='bottom')
ax_scr.set_xlim(x_lim)
ax_scr.set_ylim(x_lim)
# add score values to each box
for ai in range(n_ant):
for aj in range(n_ant):
if np.isnan(ant_scr[ai, aj]):
continue
ant_clr = np.round(img_h.cmap(img_h.norm(ant_scr[ai, aj])), decimals=2)
if np.array_equal(ant_clr, cbar_edge[0]) or np.array_equal(ant_clr, cbar_edge[1]):
txt_clr = '#ffffff'
else:
txt_clr = '#000000'
ax_scr.text(aj + 0.5, ai + 0.5, '{:+0.1f}'.format(ant_scr[ai, aj]), color=txt_clr,
horizontalalignment='center', verticalalignment='center', fontsize=12)
# final adjustments
ax_scr.set_xticks(np.arange(n_ant) + 0.5)
ax_scr.set_yticks(np.arange(n_ant) + 0.5)
ax_scr.set_xticklabels(x_tick_lbl)
ax_scr.set_yticklabels(ant_name_lst)
ax_scr.set_xlabel('Selected SOIs')
ax_scr.set_title('Association matrix from {:s}\n'.format(run_id) +
'#read (#roiFrg>{:d}, ex. vp)={:,d}, '.format(min_n_frg - 1, n_read) +
'bin-w={:d}; #perm={:d}'.format(config_lst[0]['bin_width'], n_perm)
)
plt.savefig(config_lst[0]['output_file'], bbox_inches='tight')
def perform_at_across_roi(config_lst, min_n_frg=2, n_perm=1000):
import platform
import matplotlib
if platform.system() == 'Linux':
matplotlib.use('Agg')
from matplotlib import pyplot as plt, patches
from matplotlib.colors import LinearSegmentedColormap
from utilities import load_mc4c, load_annotation, hasOL, flatten, limit_to_roi
# initialization
run_id = ','.join([config['run_id'] for config in config_lst])
configs = config_lst[0]
if configs['output_file'] is None:
configs['output_file'] = configs['output_dir'] + '/analysis_atAcrossROI_{:s}.pdf'.format(run_id)
# create bin list
edge_lst = np.linspace(configs['roi_start'], configs['roi_end'], num=201, dtype=np.int64).reshape(-1, 1)
bin_bnd = np.hstack([edge_lst[:-1], edge_lst[1:] - 1])
bin_w = bin_bnd[0, 1] - bin_bnd[0, 0]
n_bin = bin_bnd.shape[0]
# make block list
bin_cen = np.mean(bin_bnd, axis=1, dtype=np.int64).reshape(-1, 1)
# blk_crd = np.hstack([np.repeat(configs['vp_cnum'], n_bin / 3).reshape(-1, 1), edge_lst[:-3:3], edge_lst[3::3] - 1])
blk_crd = np.hstack([np.repeat(configs['vp_cnum'], n_bin).reshape(-1, 1), bin_cen - int(bin_w * 1.5), bin_cen + int(bin_w * 1.5) - 1])
blk_w = blk_crd[0, 2] - blk_crd[0, 1]
n_blk = blk_crd.shape[0]
del edge_lst
# define areas
roi_cen = np.mean([np.min(configs['prm_start']), np.max(configs['prm_end'])], dtype=np.int)
vp_crd = np.array([configs['vp_cnum'], roi_cen - int(bin_w * 1.5), roi_cen + int(bin_w * 1.5)])
roi_crd = [configs['vp_cnum'], configs['roi_start'], configs['roi_end']]
# load MC-HC data
frg_dp = load_mc4c(config_lst, unique_only=True, valid_only=True, min_mq=20, reindex_reads=True, verbose=True)
read_all = frg_dp[['ReadID', 'Chr', 'ExtStart', 'ExtEnd']].values
del frg_dp
# select >2 roi-fragments
read_inf = limit_to_roi(read_all[:, :4], vp_crd=vp_crd, roi_crd=roi_crd, min_n_frg=min_n_frg)
del read_all
# re-index reads
read_inf[:, 0] = np.unique(read_inf[:, 0], return_inverse=True)[1] + 1
n_read = len(np.unique(read_inf[:, 0]))
# convert fragments to bin-coverage
print 'Mapping reads to bins ...'
cfb_lst = [list() for i in range(n_read + 1)]
n_frg = read_inf.shape[0]
for fi in range(n_frg):
bin_idx = np.where(hasOL(read_inf[fi, 2:4], bin_bnd))[0]
cfb_lst[read_inf[fi, 0]].append(bin_idx.tolist())
# filter circles for (>1 bin cvg)
'Selecting only reads with >1 bins covered'
valid_lst = []
for rd_nid in range(1, n_read + 1):
fb_lst = cfb_lst[rd_nid]
bin_cvg = np.unique(flatten(fb_lst))
if len(bin_cvg) > 1:
valid_lst.append(rd_nid)
read_inf = read_inf[np.isin(read_inf[:, 0], valid_lst), :]
# subsample reads
# rnd_ids = np.random.choice(np.unique(read_inf[:, 0]), 6870, replace=False)
# read_inf = read_inf[np.isin(read_inf[:, 0], rnd_ids), :]
# reindexing reads
read_inf[:, 0] = np.unique(read_inf[:, 0], return_inverse=True)[1] + 1
n_read = np.max(read_inf[:, 0])
print '{:,d} reads are left after bin-coverage filter.'.format(n_read)
# get soi info
ant_pd = load_annotation(configs['genome_build'], roi_crd=roi_crd)
ant_bnd = np.hstack([ant_pd[['ant_pos']].values, ant_pd[['ant_pos']].values])
# compute score for annotations
print 'Computing expected profile for {:d} blocks (required coverage: {:d} reads):'.format(n_blk, MIN_N_POS)
blk_scr = np.full([n_blk, n_blk], fill_value=np.nan)
# x_tick_lbl = [' '] * n_blk
y_tick_lbl = [' '] * n_blk
n_ignored = 0
for bi in range(n_blk):
showprogress(bi, n_blk, n_step=20)
# add axes labels
ant_idx = np.where(hasOL(blk_crd[bi, 1:], ant_bnd, offset=0))[0]
if len(ant_idx) > 0:
ant_name = ','.join([ant_pd.loc[i, 'ant_name'] for i in ant_idx])
# x_tick_lbl[bi] = ('{:s}, #{:0.0f}'.format(ant_name, n_pos))
y_tick_lbl[bi] = ant_name
# else:
# x_tick_lbl[bi] = ('#{:0.0f}'.format(n_pos))
# ignore if vp
if hasOL(blk_crd[bi, :], vp_crd, offset=blk_w)[0]:
continue
# compute the observe and background
blk_obs, blk_rnd, read_pos = compute_mc_associations(read_inf, blk_crd[bi, :], blk_crd[:, 1:],
n_perm=n_perm, verbose=False)[:3]
n_pos = len(np.unique(read_pos[:, 0]))
if n_pos < MIN_N_POS:
n_ignored += 1
continue
# compute the scores
blk_exp = np.mean(blk_rnd, axis=0)
blk_std = np.std(blk_rnd, axis=0, ddof=0)
np.seterr(all='ignore')
blk_scr[:, bi] = np.divide(blk_obs - blk_exp, blk_std)
np.seterr(all=None)
# remove scores overlapping with positive set
is_nei = hasOL(blk_crd[bi, 1:], blk_crd[:, 1:], offset=blk_w)
blk_scr[is_nei, bi] = np.nan
if n_ignored != 0:
print '[w] {:d}/{:d} blocks are ignored due to low coverage.'.format(n_ignored, n_blk)
# set self scores to nan
# np.fill_diagonal(blk_scr, val=np.nan)
# clean up tick labels
# plotting the scores
plt.figure(figsize=(15, 13))
ax_scr = plt.subplot2grid((40, 40), (0, 0), rowspan=39, colspan=39)
ax_cmp = plt.subplot2grid((40, 40), (0, 39), rowspan=20, colspan=1)
# set up color bar
c_lim = [-6, 6]
clr_lst = ['#ff1a1a', '#ff7575', '#ffcccc', '#ffffff', '#ffffff', '#ffffff', '#ccdfff', '#3d84ff', '#3900f5']
clr_map = LinearSegmentedColormap.from_list('test', clr_lst, N=9)
clr_map.set_bad('gray', 0.1)
norm = matplotlib.colors.Normalize(vmin=c_lim[0], vmax=c_lim[1])
cbar_h = matplotlib.colorbar.ColorbarBase(ax_cmp, cmap=clr_map, norm=norm)
# cbar_h.ax.tick_params(labelsize=12)
cbar_h.ax.set_ylabel('z-score', rotation=90)
cbar_edge = np.round(cbar_h.cmap(norm(c_lim)), decimals=2)
# add score scatter matrix
x_lim = [0, n_blk]
ax_scr.imshow(blk_scr, extent=x_lim + x_lim, cmap=clr_map,
vmin=c_lim[0], vmax=c_lim[1], interpolation='nearest', origin='bottom')
ax_scr.set_xlim(x_lim)
ax_scr.set_ylim(x_lim)
# add vp patches
vp_idx = np.where(hasOL(vp_crd, blk_crd, offset=blk_w))[0]
ax_scr.add_patch(patches.Rectangle([0, vp_idx[0]], n_blk, vp_idx[-1] - vp_idx[0],
linewidth=0, edgecolor='None', facecolor='orange'))
ax_scr.add_patch(patches.Rectangle([vp_idx[0], 0], vp_idx[-1] - vp_idx[0], n_blk,
linewidth=0, edgecolor='None', facecolor='orange'))
# add score values to each box
# for bi in range(n_blk):
# for bj in range(n_blk):
# if np.isnan(blk_scr[bi, bj]):
# continue
# ant_clr = np.round(img_h.cmap(img_h.norm(blk_scr[bi, bj])), decimals=2)
# if np.array_equal(ant_clr, cbar_edge[0]) or np.array_equal(ant_clr, cbar_edge[1]):
# txt_clr = '#ffffff'
# else:
# txt_clr = '#000000'
# ax_scr.text(bj + 0.5, bi + 0.5, '{:+0.1f}'.format(blk_scr[bi, bj]), color=txt_clr,
# horizontalalignment='center', verticalalignment='center', fontsize=12)
# adjust ticks
for lbl in np.unique(y_tick_lbl):
if lbl == ' ':
continue
idx_lst = np.where(np.isin(y_tick_lbl, lbl))[0]
if len(idx_lst) > 1:
kpt_idx = np.mean(idx_lst, dtype=np.int)
for idx in idx_lst:
y_tick_lbl[idx] = 'l'
y_tick_lbl[kpt_idx] = lbl + ' '
# final adjustments
ax_scr.set_xticks(np.arange(n_blk) + 0.5)
ax_scr.set_yticks(np.arange(n_blk) + 0.5)
ax_scr.set_xticklabels(y_tick_lbl, rotation=90)
ax_scr.set_yticklabels(y_tick_lbl)
ax_scr.set_xlabel('Selected SOIs')
ax_scr.set_title('Association matrix from {:s}\n'.format(configs['run_id']) +
'#read (#roiFrg>{:d}, ex. vp)={:,d}, '.format(min_n_frg - 1, n_read) +
'bin-w={:0.0f}; block-w={:0.0f}; #perm={:d}'.format(bin_w, blk_w, n_perm)
)
plt.savefig(configs['output_file'], bbox_inches='tight')