update obo_extractor

Author: sfluegel05

.gitignore

@@ -205,3 +205,7 @@ cython_debug/
 marimo/_static/
 marimo/_lsp/
 __marimo__/
+
+
+
+data/

chebi_utils/obo_extractor.py

@@ -22,7 +22,7 @@ def _term_data(doc: "fastobo.term.TermFrame") -> dict | None:
         Parsed term data, or ``None`` if the term is marked as obsolete.
     """
     parents: list[str] = []
-    has_part: set[str] = set()
+    relations: dict = dict()
     name: str | None = None
     smiles: str | None = None
     subset: str | None = None
@@ -42,8 +42,10 @@ def _term_data(doc: "fastobo.term.TermFrame") -> dict | None:
             if "SMILES" in clause.raw_value() and smiles is None:
                 smiles = clause.raw_value().split('"')[1]
         elif isinstance(clause, fastobo.term.RelationshipClause):
-            if str(clause.typedef) == "has_part":
-                has_part.add(_chebi_id_to_str(str(clause.term)))
+            relation = str(clause.typedef)
+            if relation not in relations:
+                relations[relation] = []
+            relations[relation].append(_chebi_id_to_str(str(clause.term)))
         elif isinstance(clause, fastobo.term.IsAClause):
             parents.append(_chebi_id_to_str(str(clause.term)))
         elif isinstance(clause, fastobo.term.NameClause):
@@ -54,7 +56,7 @@ def _term_data(doc: "fastobo.term.TermFrame") -> dict | None:
     return {
         "id": _chebi_id_to_str(str(doc.id)),
         "parents": parents,
-        "has_part": has_part,
+        "relations": relations,
         "name": name,
         "smiles": smiles,
         "subset": subset,
@@ -107,7 +109,14 @@ def build_chebi_graph(filepath: str | Path) -> nx.DiGraph:
         for parent_id in term["parents"]:
             graph.add_edge(node_id, parent_id, relation="is_a")
 
-        for part_id in term["has_part"]:
-            graph.add_edge(node_id, part_id, relation="has_part")
+        for relation, parts in term["relations"].items():
+            for part_id in parts:
+                graph.add_edge(node_id, part_id, relation=relation)
 
     return graph
+
+def get_hierarchy_subgraph(chebi_graph: nx.DiGraph) -> nx.DiGraph:
+    """Subgraph of ChEBI including only edges corresponding to hierarchical relations (is_a). Also removes nodes that are not connected by any is_a edges to other nodes."""
+    return chebi_graph.edge_subgraph(
+        (u, v) for u, v, d in chebi_graph.edges(data=True) if d.get("relation") == "is_a"
+    )

chebi_utils/sdf_extractor.py

@@ -10,10 +10,21 @@
 from rdkit import Chem
 
 
-def _update_mol_valences(mol: Chem.Mol) -> Chem.Mol:
-    """Mark all atoms as having no implicit hydrogens to preserve molfile valences."""
-    for atom in mol.GetAtoms():
-        atom.SetNoImplicit(True)
+def _sanitize_molecule(mol: Chem.Mol) -> Chem.Mol:
+    """Sanitize molecule, mirroring the ChEBI molecule processing."""
+    from chembl_structure_pipeline.standardizer import update_mol_valences
+
+    mol = update_mol_valences(mol)
+    Chem.SanitizeMol(
+        mol,
+        sanitizeOps=Chem.SanitizeFlags.SANITIZE_FINDRADICALS
+        | Chem.SanitizeFlags.SANITIZE_KEKULIZE
+        | Chem.SanitizeFlags.SANITIZE_SETAROMATICITY
+        | Chem.SanitizeFlags.SANITIZE_SETCONJUGATION
+        | Chem.SanitizeFlags.SANITIZE_SETHYBRIDIZATION
+        | Chem.SanitizeFlags.SANITIZE_SYMMRINGS,
+        catchErrors=True,
+    )
     return mol
 
 
@@ -39,19 +50,7 @@ def _parse_molblock(molblock: str, chebi_id: str | None = None) -> Chem.Mol | No
     if mol is None:
         warnings.warn(f"Failed to parse molblock for {chebi_id}", stacklevel=2)
         return None
-    mol = _update_mol_valences(mol)
-    Chem.SanitizeMol(
-        mol,
-        sanitizeOps=(
-            Chem.SanitizeFlags.SANITIZE_FINDRADICALS
-            | Chem.SanitizeFlags.SANITIZE_KEKULIZE
-            | Chem.SanitizeFlags.SANITIZE_SETAROMATICITY
-            | Chem.SanitizeFlags.SANITIZE_SETCONJUGATION
-            | Chem.SanitizeFlags.SANITIZE_SETHYBRIDIZATION
-            | Chem.SanitizeFlags.SANITIZE_SYMMRINGS
-        ),
-        catchErrors=True,
-    )
+    mol = _sanitize_molecule(mol)
     return mol
 
 

pyproject.toml

@@ -15,6 +15,7 @@ dependencies = [
     "numpy>=1.24",
     "pandas>=2.0",
     "rdkit>=2022.09",
+    "chembl_structure_pipeline>=1.2.4",
 ]
 
 [project.optional-dependencies]

tests/fixtures/sample.obo

@@ -1,37 +1,87 @@
 format-version: 1.2
+date: 06:01:2026 14:33
+saved-by: chebi
+subsetdef: 1:STAR "Preliminary entries"
+subsetdef: 2:STAR "Annotated by 3rd party"
+subsetdef: 3:STAR "Manually annotated by ChEBI Team"
+synonymtypedef: BRAND:NAME "BRAND NAME"
+synonymtypedef: http://purl.obolibrary.org/obo/chebi/INN "INN"
+synonymtypedef: IUPAC:NAME "IUPAC NAME"
+default-namespace: chebi_ontology
+idspace: chemrof https://w3id.org/chemrof/ 
+idspace: foaf http://xmlns.com/foaf/0.1/ 
+idspace: oboInOwl http://www.geneontology.org/formats/oboInOwl# 
+idspace: terms http://purl.org/dc/terms/ 
+remark: ChEBI subsumes and replaces the Chemical Ontology first. This ontology was developed by Michael Ashburner and Pankaj Jaiswal. Data was cuarated by The ChEBI Curation Team. For any queries contact chebi-help@ebi.ac.uk
 ontology: chebi
+property_value: foaf:homepage "https://www.ebi.ac.uk/chebi" xsd:anyURI
+property_value: owl:versionInfo "248" xsd:decimal
+property_value: terms:description "Chemical Entities of Biological Interest, also known as ChEBI, is a chemical database and ontology of molecular entities focused on 'small' chemical compounds, that is part of the Open Biomedical Ontologies (OBO) effort at the European Bioinformatics Institute (EBI)." xsd:string
+property_value: terms:license https://creativecommons.org/licenses/by/4.0/
+property_value: terms:title "ChEBI Ontology" xsd:string
 
 [Term]
-id: CHEBI:1
-name: compound A
-def: "A test compound." [TestDB:001]
-is_a: CHEBI:2 ! compound B
-relationship: has_part CHEBI:3 ! methyl group
-property_value: http://purl.obolibrary.org/obo/chebi/smiles "C" xsd:string
+id: CHEBI:10
+name: (+)-Atherospermoline
+subset: 2:STAR
+synonym: "(+)-Atherospermoline" RELATED [kegg.compound]
+xref: cas:21008-67-3 {source="cas"}
+xref: kegg.compound:C11141 {source="kegg.compound"}
+xref: knapsack:C00001814 {source="knapsack"}
+is_a: CHEBI:133004 ! bisbenzylisoquinoline alkaloid
+property_value: chemrof:charge "0" xsd:integer
+property_value: chemrof:generalized_empirical_formula "C36H38N2O6" xsd:string
+property_value: chemrof:inchi_key_string "XGEAUXVPBXUBKN-NSOVKSMOSA-N" xsd:string
+property_value: chemrof:inchi_string "InChI=1S/C36H38N2O6/c1-37-13-11-23-18-31(41-3)32-20-26(23)27(37)15-21-5-8-25(9-6-21)43-30-17-22(7-10-29(30)39)16-28-34-24(12-14-38(28)2)19-33(42-4)35(40)36(34)44-32/h5-10,17-20,27-28,39-40H,11-16H2,1-4H3/t27-,28-/m0/s1" xsd:string
+property_value: chemrof:mass "594.708" xsd:decimal
+property_value: chemrof:monoisotopic_mass "594.27299" xsd:decimal
+property_value: chemrof:smiles_string "COc1cc2c3cc1Oc1c(O)c(OC)cc4c1[C@H](Cc1ccc(O)c(c1)Oc1ccc(cc1)C[C@@H]3N(C)CC2)N(C)CC4" xsd:string
 
 [Term]
-id: CHEBI:2
-name: compound B
-def: "Another test compound." [TestDB:002]
-is_a: CHEBI:5 ! root compound
+id: CHEBI:133004
+name: bisbenzylisoquinoline alkaloid
+def: "A type of benzylisoquinoline alkaloid whose structures are built up of two benzylisoquinoline units linked by ether bridges. Various structural patterns resulting from additional bridging between the two units by direct carbon-carbon bridging or by methylenedioxy groups are common." []
+subset: 3:STAR
+synonym: "bis(benzylisoquinoline) alkaloid" RELATED [chebi]
+synonym: "bis(benzylisoquinoline) alkaloids" RELATED [chebi]
+synonym: "bis-benzylisoquinoline alkaloid" RELATED [chebi]
+synonym: "bis-benzylisoquinoline alkaloids" RELATED [chebi]
+synonym: "bisbenzylisoquinoline alkaloids" RELATED [chebi]
+xref: pubmed:1955879 {source="pubmed"}
+xref: pubmed:2191354 {source="pubmed"}
+xref: pubmed:3323421 {source="pubmed"}
+is_a: CHEBI:22750 ! benzylisoquinoline alkaloid
 
 [Term]
-id: CHEBI:3
-name: methyl group
-def: "A methyl group." [TestDB:003]
-subset: 3_STAR
+id: CHEBI:22750
+name: benzylisoquinoline alkaloid
+def: "Any isoquinoline alkaloid based on a benzylisoquinoline skeleton." []
+subset: 3:STAR
+synonym: "benzylisoquinoline alkaloids" RELATED [chebi]
+is_a: CHEBI:24921 ! isoquinoline alkaloid
 
 [Term]
-id: CHEBI:4
-name: obsolete term
-def: "This term is obsolete." [TestDB:004]
+id: CHEBI:27189
 is_obsolete: true
 
 [Term]
-id: CHEBI:5
-name: root compound
-def: "The root compound." [TestDB:005]
-
-[Typedef]
-id: has_part
-name: has part
+id: CHEBI:28017
+name: starch
+alt_id: CHEBI:26750
+alt_id: CHEBI:26751
+alt_id: CHEBI:9251
+def: "The most important reserve polysaccharide found in plants. It is a glucan consisting of amylose and amylopectin." []
+subset: 3:STAR
+synonym: "amidon" RELATED [chebi]
+synonym: "amylum" RELATED [chebi]
+synonym: "Staerke" RELATED [chebi]
+synonym: "Starch" RELATED [kegg.compound]
+xref: cas:9005-25-8 {source="cas"}
+xref: kegg.compound:C00369 {source="kegg.compound"}
+xref: kegg.drug:D06507 {source="kegg.drug"}
+xref: kegg.glycan:G10545 {source="kegg.glycan"}
+xref: wikipedia.en:Starch {source="wikipedia.en"}
+is_a: CHEBI:37163 ! glucan
+relationship: BFO:0000051 CHEBI:28057 ! has part amylopectin
+relationship: BFO:0000051 CHEBI:28102 ! has part amylose
+relationship: RO:0000087 CHEBI:75771 ! has role mouse metabolite

tests/test_obo_extractor.py

@@ -18,7 +18,7 @@ def test_returns_directed_graph(self):
         assert isinstance(g, nx.DiGraph)
 
     def test_correct_number_of_nodes(self):
-        # CHEBI:4 is obsolete and must be excluded -> 4 nodes remain
+        # CHEBI:27189 is obsolete -> excluded; 3 explicit + 1 implicit (24921) = 4
         g = build_chebi_graph(SAMPLE_OBO)
         assert len(g.nodes) == 4
 
@@ -28,55 +28,50 @@ def test_node_ids_are_strings(self):
 
     def test_expected_nodes_present(self):
         g = build_chebi_graph(SAMPLE_OBO)
-        assert set(g.nodes) == {"1", "2", "3", "5"}
+        assert set(g.nodes) == {"10", "133004", "22750", "24921"}
 
     def test_obsolete_term_excluded(self):
         g = build_chebi_graph(SAMPLE_OBO)
-        assert "4" not in g.nodes
+        assert "27189" not in g.nodes
 
     def test_node_name_attribute(self):
         g = build_chebi_graph(SAMPLE_OBO)
-        assert g.nodes["1"]["name"] == "compound A"
-        assert g.nodes["2"]["name"] == "compound B"
+        assert g.nodes["10"]["name"] == "(+)-Atherospermoline"
+        assert g.nodes["133004"]["name"] == "bisbenzylisoquinoline alkaloid"
 
     def test_smiles_extracted_from_property_value(self):
         g = build_chebi_graph(SAMPLE_OBO)
-        assert g.nodes["1"]["smiles"] == "C"
+        expected = "COc1cc2c3cc1Oc1c(O)c(OC)cc4c1[C@H](Cc1ccc(O)c(c1)Oc1ccc(cc1)C[C@@H]3N(C)CC2)N(C)CC4"
+        assert g.nodes["10"]["smiles"] == expected
 
     def test_smiles_none_when_absent(self):
         g = build_chebi_graph(SAMPLE_OBO)
-        assert g.nodes["2"]["smiles"] is None
+        assert g.nodes["133004"]["smiles"] is None
 
     def test_subset_extracted(self):
         g = build_chebi_graph(SAMPLE_OBO)
-        assert g.nodes["3"]["subset"] == "3_STAR"
-
-    def test_subset_none_when_absent(self):
-        g = build_chebi_graph(SAMPLE_OBO)
-        assert g.nodes["1"]["subset"] is None
+        assert g.nodes["133004"]["subset"] == "3:STAR"
 
     def test_isa_edge_present(self):
         g = build_chebi_graph(SAMPLE_OBO)
-        # CHEBI:1 is_a CHEBI:2
-        assert g.has_edge("1", "2")
-        assert g.edges["1", "2"]["relation"] == "is_a"
+        # CHEBI:10 is_a CHEBI:133004
+        assert g.has_edge("10", "133004")
+        assert g.edges["10", "133004"]["relation"] == "is_a"
 
-    def test_has_part_edge_present(self):
+    def test_isa_chain(self):
         g = build_chebi_graph(SAMPLE_OBO)
-        # CHEBI:1 has_part CHEBI:3
-        assert g.has_edge("1", "3")
-        assert g.edges["1", "3"]["relation"] == "has_part"
+        # CHEBI:133004 is_a CHEBI:22750
+        assert g.has_edge("133004", "22750")
+        assert g.edges["133004", "22750"]["relation"] == "is_a"
+        # CHEBI:22750 is_a CHEBI:24921
+        assert g.has_edge("22750", "24921")
+        assert g.edges["22750", "24921"]["relation"] == "is_a"
 
     def test_total_edge_count(self):
         g = build_chebi_graph(SAMPLE_OBO)
-        # 1->2 (is_a), 1->3 (has_part), 2->5 (is_a)
+        # 10->133004 (is_a), 133004->22750 (is_a), 22750->24921 (is_a)
         assert len(g.edges) == 3
 
-    def test_typedef_stanza_excluded(self):
-        g = build_chebi_graph(SAMPLE_OBO)
-        # "has_part" Typedef id is not numeric CHEBI ID, should not appear as node
-        assert "has_part" not in g.nodes
-
     def test_xref_lines_do_not_break_parsing(self, tmp_path):
         obo_with_xrefs = tmp_path / "xref.obo"
         obo_with_xrefs.write_text(