fixed gene_adata obs table inference when adding from adata

fairliereese · fairliereese · commit bacc2f4eeb96 · 2023-07-05T21:50:32.000-07:00
diff --git a/swan_vis/swangraph.py b/swan_vis/swangraph.py
@@ -10,7 +10,6 @@
 import sqlite3
 import pickle
 import anndata
-import diffxpy.api as de
 from statsmodels.stats.multitest import multipletests
 import multiprocessing
 from itertools import repeat
@@ -336,14 +335,30 @@ def abundance_to_adata(self,
 		df = df.T
 
 		# get adata components - obs, var, and X
+
+		# var
 		var = df.columns.to_frame()
 		var.columns = [id_col]
 		var.index.name = 'tid'
+
+		# obs
 		obs = df.index.to_frame()
 		obs.columns = ['dataset']
+
+		# if we already have transcript abundance and we're adding genes,
+		# copy the obs information there
+		if how == 'gene' and self.has_abundance():
+			self.adata.obs = reset_dupe_index(self.adata.obs, 'dataset')
+			obs = obs.merge(self.adata.obs, how='left', on='dataset')
+			obs.drop('dataset_back', axis=1, inplace=True)
+			self.adata.obs = set_dupe_index(self.adata.obs, 'dataset')
+
 		obs.index.name = 'dataset'
+
+		# X
 		X = sparse.csr_matrix(df.to_numpy())
 
+
 		# create transcript-level adata object and filter out unexpressed transcripts
 		adata = anndata.AnnData(var=var, obs=obs, X=X)
 		genes, _  = sc.pp.filter_genes(adata, min_counts=1, inplace=False)
