fix: allow non-digit valid gene names and clean product descriptions independently of name (fixes #90)

funannotate2/name_cleaner.py

@@ -381,15 +381,16 @@ def process_annotation(self, annotation, gene_id=None):
                         del result["name"]
 
         # Process product if present
-        if "product" in result and "name" in result:
+        if "product" in result:
             products = result["product"]
-            names = result["name"]
+            names = result.get("name", [])
+            name = names[0] if (isinstance(names, list) and names) else None
 
-            if isinstance(products, list) and products and isinstance(names, list) and names:
+            if isinstance(products, list) and products:
                 # If we have a curated product, it's already set above
-                if not self.get_curated_product(names[0]):
+                if not name or not self.get_curated_product(name):
                     # Clean the product
-                    cleaned_product = self.clean_product(products[0], names[0])
+                    cleaned_product = self.clean_product(products[0], name)
                     result["product"] = [cleaned_product]
 
         return result

funannotate2/search.py

@@ -735,9 +735,12 @@ def add2dict(adict, gene, key, value):
 
 
 def swissprot_valid_gene(name):
+    if not name:
+        return False
     if (
-        number_present(name)
-        and len(name) > 2
+        len(name) > 2
+        and not name.startswith("orf")
+        and not name[0].isdigit()
         and not morethanXnumbers(name, 3)
         and "." not in name
     ):

tests/unit/test_search_comprehensive.py

@@ -343,8 +343,11 @@ def test_swissprot2tsv(self, mock_json_dump):
         assert "seq1" in result
         assert "db_xref" in result["seq1"]
         assert result["seq1"]["db_xref"] == ["UniProtKB/Swiss-Prot:P12345"]
-        assert "note" in result["seq1"]
-        assert "80.0% identical to TEST_HUMAN Test protein" in result["seq1"]["note"][0]
+        assert "name" in result["seq1"]
+        assert result["seq1"]["name"] == ["TEST"]
+        assert "product" in result["seq1"]
+        assert result["seq1"]["product"] == ["Test protein"]
+        assert "note" not in result["seq1"]
         assert mock_json_dump.call_count == 1
 
     @patch("funannotate2.search.os.path.isdir")