Issue32

bin/scdrs

@@ -167,6 +167,21 @@ def compute_score(
     )
     print("First 3 cells: %s" % (str(list(adata.obs_names[:3]))))
     print("First 5 genes: %s" % (str(list(adata.var_names[:5]))))
+    if ADJ_PROP is not None:
+        err_msg = "'adj_prop'=%s not in 'adata.obs.columns'" % ADJ_PROP
+        assert ADJ_PROP in adata.obs, err_msg
+        temp_df = adata.obs[[ADJ_PROP]].groupby(ADJ_PROP).agg({ADJ_PROP: len})
+        temp_df.columns = ["n"]
+        temp_df.sort_values(by="n", ascending=False, inplace=True)
+        print("--adj-prop: %d categories" % temp_df.shape[0])
+        msg = ", ".join(
+            ["%s (%d)" % (x, temp_df.loc[x, "n"]) for x in temp_df.index[:3]]
+        )
+        print("Top 3: %s" % msg)
+        msg = ", ".join(
+            ["%s (%d)" % (x, temp_df.loc[x, "n"]) for x in temp_df.index[-3:]]
+        )
+        print("Bottom 3: %s" % msg)
 
     # Load .cov file
     if COV_FILE is not None:

scdrs/pp.py

@@ -4,6 +4,7 @@
 import pandas as pd
 from skmisc.loess import loess
 from typing import List
+import warnings
 
 
 def category2dummy(
@@ -373,6 +374,10 @@ def compute_stats(
 
     # Borrowed from scanpy _highly_variable_genes_seurat_v3
     not_const = df_gene["ct_var"].values > 0
+    if (df_gene["ct_mean"].values <= 0).sum() > 0:
+        # Exclude genes with negative values (usually small)
+        warnings.warn("%d genes with ct_mean<0" % (df_gene["ct_mean"].values < 0).sum())
+        not_const = not_const & (df_gene["ct_mean"].values > 0)
     estimat_var = np.zeros(adata.shape[1], dtype=np.float64)
     y = np.log10(df_gene["ct_var"].values[not_const])
     x = np.log10(df_gene["ct_mean"].values[not_const])

scdrs/version.py

@@ -2,5 +2,5 @@
 # 1) we don't load dependencies by storing it in __init__.py
 # 2) we can import it in setup.py for the same reason
 # 3) we can import it into your module module
-__version__ = '1.0.2'
+__version__ = '1.0.3'
 __version_info__ = tuple([ int(num) for num in __version__.split('.')])