fix: Remove empty data before fitting

.gitignore

@@ -2,6 +2,7 @@ data/
 data*/
 real-data*/
 old-real-data*/
+performance/
 raw_data/
 !raw_data/alignment.bam
 !raw_data/names-mdmg.dmp

src/metaDMG/fit/fits.py

@@ -552,6 +552,19 @@ def filter_k_sum(
     return df_mismatches.query("k_sum_total > 0")
 
 
+def filter_max_N_in_group(
+    config: Config,
+    df_mismatches: pd.DataFrame,
+) -> pd.DataFrame:
+
+    # filter out tax_id's with 0 k_sum_total
+    tax_ids_to_drop = set(df_mismatches.query("max_N_in_group == 0")["tax_id"].unique())
+    if len(tax_ids_to_drop) > 0:
+        logger.debug(f"Dropping the following Tax IDs since max_N_in_group == 0:")
+        logger.debug(tax_ids_to_drop)
+    return df_mismatches.query("max_N_in_group > 0")
+
+
 #%%
 
 
@@ -569,6 +582,15 @@ def compute(config, df_mismatches):
         logger.debug(s)
         raise BadDataError(s)
 
+    # filter out tax_id's with no data in them (max_N_in_group == 0)
+    df_mismatches = filter_max_N_in_group(config, df_mismatches)
+
+    if len(df_mismatches) == 0:
+        s = f"{config['sample']} df_mismatches.query('max_N_in_group > 0') is empty."
+        logger.debug("WARNING: BadDataError")
+        logger.debug(s)
+        raise BadDataError(s)
+
     # # filter out tax_id's with 0 k_sum_total
     # df_mismatches = filter_k_sum(config, df_mismatches)
 

src/metaDMG/fit/mismatches.py

@@ -148,18 +148,37 @@ def add_k_sum_counts(df):
     return df
 
 
-def compute_min_N_in_group(group, config):
-    if config["forward_only"]:
-        return group[group.position > 0][bases_forward[0]].min()
-    else:
-        min_N_forward = group[group.position > 0][bases_forward[0]].min()
-        min_N_reverse = group[group.position < 0][bases_reverse[0]].min()
-        return min(min_N_forward, min_N_reverse)
+def compute_min_max_N_in_group(group, config):
+    min_N, max_N = group["N"].min(), group["N"].max()
+    return pd.Series({"min_N_in_group": min_N, "max_N_in_group": max_N})
+
+    # mask_forward = group.position > 0
+    # mask_reverse = group.position < 0
+
+    # if config["forward_only"]:
+    #     min_N = group[mask_forward][bases_forward[0]].min()
+    #     max_N = group[mask_forward][bases_forward[0]].max()
+    #     # return min_N, max_N
+    #     return pd.Series({"min_N_in_group": min_N, "max_N_in_group": max_N})
 
+    # min_N_forward = group[mask_forward > 0][bases_forward[0]].min()
+    # min_N_reverse = group[mask_reverse < 0][bases_reverse[0]].min()
+    # min_N = min(min_N_forward, min_N_reverse)
 
-def add_min_N_in_group(df, config):
-    ds = df.groupby("tax_id").apply(compute_min_N_in_group, config)
-    ds = ds.reset_index().rename(columns={0: "min_N_in_group"})
+    # max_N_forward = group[mask_forward > 0][bases_forward[0]].max()
+    # max_N_reverse = group[mask_reverse < 0][bases_reverse[0]].max()
+    # max_N = max(max_N_forward, max_N_reverse)
+
+    # # return min_N, max_N
+    # return pd.Series({"min_N_in_group": min_N, "max_N_in_group": max_N})
+
+
+def add_min_max_N_in_group(df, config):
+    ds = (
+        df.groupby("tax_id")
+        .apply(compute_min_max_N_in_group, config)
+        .reset_index(drop=False)
+    )
     df = pd.merge(df, ds, on=["tax_id"])
     return df
 
@@ -213,7 +232,7 @@ def compute(config: Config) -> pd.DataFrame:
         .pipe(make_reverse_position_negative)
         .pipe(add_k_N_x_names, config)
         .pipe(add_k_sum_counts)
-        .pipe(add_min_N_in_group, config)
+        .pipe(add_min_max_N_in_group, config)
         .pipe(make_tax_id_str)
         .reset_index(drop=True)
         .fillna(0)