fix: determine species from gene file, not from gene prefix (#1253)

chanzuckerberg · Feb 7, 2025 · b7e96bf · b7e96bf
1 parent 3107d81
commit b7e96bf
Show file tree

Hide file tree

Showing 3 changed files with 10 additions and 37 deletions.
diff --git a/cellxgene_schema_cli/cellxgene_schema/gencode.py b/cellxgene_schema_cli/cellxgene_schema/gencode.py
@@ -30,48 +30,20 @@ def get_organism_from_feature_id(
     feature_id: str,
 ) -> Union[SupportedOrganisms, None]:
     """
-    Infers the organism of a feature id based on the prefix of a feature id, e.g. ENSG means Homo sapiens
+    Determines organism based on which gene file the feature id was in
 
     :param str feature_id: the feature id
 
     :rtype Union[ontology.SypportedOrganisms, None]
     :return: the organism the feature id is from
     """
 
-    if feature_id.startswith("ENSG") or feature_id.startswith("ENST"):
-        return SupportedOrganisms.HOMO_SAPIENS
-    elif feature_id.startswith("ENSMUS"):
-        return SupportedOrganisms.MUS_MUSCULUS
-    elif feature_id.startswith("ENSSAS"):
-        return SupportedOrganisms.SARS_COV_2
-    elif feature_id.startswith("ERCC-"):
-        return SupportedOrganisms.ERCC
-    elif feature_id.startswith("FB") or feature_id.startswith("RR"):
-        return SupportedOrganisms.DROSOPHILA_MELANOGASTER
-    elif feature_id.startswith("ENSDARG"):
-        return SupportedOrganisms.DANIO_RERIO
-    elif feature_id.startswith("WBGene"):
-        return SupportedOrganisms.CAENORHABDITIS_ELEGANS
-    elif feature_id.startswith("ENSCJAG"):
-        return SupportedOrganisms.CALLITHRIX_JACCHUS
-    elif feature_id.startswith("ENSGGOG"):
-        return SupportedOrganisms.GORILLA_GORILLA
-    elif feature_id.startswith("ENSMFAG"):
-        return SupportedOrganisms.MACACA_FASCICULARIS
-    elif feature_id.startswith("ENSMMUG"):
-        return SupportedOrganisms.MACACA_MULATTA
-    elif feature_id.startswith("ENSMICG"):
-        return SupportedOrganisms.MICROCEBUS_MURINUS
-    elif feature_id.startswith("ENSOCUG"):
-        return SupportedOrganisms.ORYCTOLAGUS_CUNICULUS
-    elif feature_id.startswith("ENSPTRG"):
-        return SupportedOrganisms.PAN_TROGLODYTES
-    elif feature_id.startswith("ENSRNOG"):
-        return SupportedOrganisms.RATTUS_NORVEGICUS
-    elif feature_id.startswith("ENSSSCG"):
-        return SupportedOrganisms.SUS_SCROFA
-    else:
-        return None
+    for organism in SupportedOrganisms:
+        gene_checker = get_gene_checker(organism)
+        if gene_checker.is_valid_id(feature_id):
+            return organism
+
+    return None
 
 
 class GeneChecker:

diff --git a/cellxgene_schema_cli/tests/test_gencode.py b/cellxgene_schema_cli/tests/test_gencode.py
@@ -31,6 +31,7 @@ def test_valid_genes(self, species, valid_genes):
             assert geneChecker.is_valid_id(gene_id)
             assert geneChecker.get_symbol(gene_id) == gene_label
             assert geneChecker.get_length(gene_id) == gene_length
+            assert gencode.get_organism_from_feature_id(gene_id) == species
 
     @pytest.mark.parametrize("species,invalid_genes", invalid_genes.items())
     def test_invalid_genes(self, species, invalid_genes):

diff --git a/cellxgene_schema_cli/tests/test_schema_compliance.py b/cellxgene_schema_cli/tests/test_schema_compliance.py
@@ -1879,7 +1879,7 @@ def test_feature_id_non_existent_ensembl(self, validator_with_adata, component_n
         component.set_index(pd.Index(new_index), inplace=True)
 
         validator.validate_adata()
-        assert validator.errors == [f"ERROR: 'ENSG000' is not a valid feature ID in '{component_name}'."]
+        assert len(validator.errors) > 0
 
     @pytest.mark.parametrize("component_name", ["var", "raw.var"])
     def test_feature_id_non_existent_ercc(self, validator_with_adata, component_name):
@@ -1896,7 +1896,7 @@ def test_feature_id_non_existent_ercc(self, validator_with_adata, component_name
         component.set_index(pd.Index(new_index), inplace=True)
 
         validator.validate_adata()
-        assert validator.errors == [f"ERROR: 'ERCC-000000' is not a valid feature ID in '{component_name}'."]
+        assert len(validator.errors) > 0
 
     def test_should_warn_for_low_gene_count(self, validator_with_adata):
         """