added pd. NA case in test_lovd_fill_hg38.py, added mask for values in…

… lovd_fill_hg38

api/data/refactoring.py

@@ -192,30 +192,34 @@ def lovd_fill_hg38(lovd: pd.DataFrame):
 
     if lovd.empty:
         return
-    lovd['hg38_gnomad_format'] = lovd['VariantOnGenome/DNA/hg38'].replace('', pd.NA)
-    missing_hg38_mask = lovd['hg38_gnomad_format'].isna()
+    lovd.loc[:,'hg38_gnomad_format'] = lovd.loc[:,'VariantOnGenome/DNA/hg38'].replace('', pd.NA)
+    missing_hg38_mask = lovd.loc[:,'hg38_gnomad_format'].isna()
     lovd.loc[missing_hg38_mask, 'hg38_gnomad_format'] = lovd.loc[missing_hg38_mask, 'VariantOnGenome/DNA'].apply(
         convert_hg19_if_missing)
-    lovd['hg38_gnomad_format'] = lovd['hg38_gnomad_format'].apply(convert_to_gnomad_gen)
+    lovd.loc[:,'hg38_gnomad_format'] = lovd.loc[:,'hg38_gnomad_format'].apply(convert_to_gnomad_gen)
 
 
-def convert_hg19_if_missing(hg19: pd.Series, lo = LiftOver('hg19', 'hg38')):
+def convert_hg19_if_missing(hg19: str, lo = LiftOver('hg19', 'hg38')):
     """
     Converts hg19 variant to hg38 if hg38 is missing.
     :param hg19: a row from the DataFrame.
     :param lo: converter for genomic data between reference assemblies
     :return: hg38 value or a conversion of the hg19 value in the format 'g.positionref>alt'.
     """
 
-    if pd.isna(hg19):
+    if pd.isna(hg19) or '_' in hg19:
         return "?"
-    if '?' in hg19 or "_" in hg19:
+
+    match = re.search(r'g\.(\d+)', hg19)
+    if not match:
         return '?'
-    position_str = hg19[2:10]
+
+    position_str = match.group(1)
     new_pos = lo.convert_coordinate('chr6', int(position_str))[0][1]
     return f"g.{new_pos}{hg19[-3:]}"
 
 
+
 def convert_to_gnomad_gen(variant: str):
     """
     converts a variant string from hg38 format

tests/pipeline.ipynb

@@ -4018,8 +4018,8 @@
    "metadata": {
     "collapsed": false,
     "ExecuteTime": {
-     "end_time": "2024-09-03T08:10:37.655840Z",
-     "start_time": "2024-09-03T08:10:28.838482Z"
+     "end_time": "2024-09-03T14:19:14.730427Z",
+     "start_time": "2024-09-03T14:19:05.969159Z"
     }
    },
    "id": "dd9b17623f26a07c",

tests/test_lovd_fill_hg38.py

@@ -88,6 +88,18 @@ def test_fill_hg38_no_variants(self):
         lovd_fill_hg38(self.df)
         self.assertEqual(self.df.shape[0], 0, "Empty dataframe should not add rows.")
 
+    def test_fill_hg38_NA_variants(self):
+        """Test filling hg38 values when there are pd. NA variants in the dataframe."""
+        self.df = pd.DataFrame({
+            'VariantOnGenome/DNA': [pd.NA],
+            'VariantOnGenome/DNA/hg38': [pd.NA]
+        })
+        lovd_fill_hg38(self.df)
+        expected_values = ['?']
+        self.assertIn('hg38_gnomad_format', self.df.columns,
+                      "Column 'hg38_gnomad_format' should be added.")
+        self.assertListEqual(self.df['hg38_gnomad_format'].tolist(), expected_values)
+
 
 
 if __name__ == '__main__':