Merge pull request #1310 from bgyori/ido

Improved OWL/IDO processing and handling of external ontology relations

indra/databases/owl_client.py

@@ -5,7 +5,7 @@
 import pickle
 from collections import defaultdict
 from operator import itemgetter
-from typing import Any, Mapping, TYPE_CHECKING
+from typing import Any, Collection, Mapping, TYPE_CHECKING
 
 from tqdm import tqdm
 
@@ -24,6 +24,7 @@ def entry_from_term(
         term: "pronto.Term",
         prefix: str,
         remove_prefix: bool = False,
+        allowed_external_ns: Collection = None,
     ) -> Mapping[str, Any]:
         """Create a data dictionary from a Pronto term."""
         rels_dict = defaultdict(list)
@@ -35,12 +36,18 @@ def entry_from_term(
                 continue
             else:
                 xrefs.append(dict(namespace=xref_db, id=xref_id))
-        for child in term.subclasses(distance=1, with_self=False):
-            child_db, child_id = child.id.split(':', maxsplit=1)
-            if remove_prefix and child_db.lower() == prefix.lower():
-                rels_dict["is_a"].append(child_id)
+        for parent in term.superclasses(distance=1, with_self=False):
+            parent_db, parent_id = parent.id.split(':', maxsplit=1)
+            # If the parent here is not from this namespace and not one of the
+            # allowed external namespaces then we skip the parent
+            if parent_db.lower() != prefix.lower() and \
+                    (not allowed_external_ns or
+                     parent_db not in allowed_external_ns):
+                continue
+            if remove_prefix and parent_db.lower() == prefix.lower():
+                rels_dict["is_a"].append(parent_id)
             else:
-                rels_dict["is_a"].append(child.id)
+                rels_dict["is_a"].append(parent.id)
 
         term_ns, term_id = term.id.split(':', maxsplit=1)
         term_ns = term_ns.lower()
@@ -62,6 +69,7 @@ def entries_from_ontology(
         *,
         skip_obsolete: bool = True,
         remove_prefix: bool = False,
+        allowed_external_ns: Collection = None,
     ):
         prefix = prefix.upper()
         rv = []

indra/ontology/bio/ontology.py

@@ -19,7 +19,7 @@ class BioOntology(IndraOntology):
     # should be incremented to "force" rebuilding the ontology to be consistent
     # with the underlying resource files.
     name = 'bio'
-    version = '1.13'
+    version = '1.14'
 
     def __init__(self):
         super().__init__()
@@ -187,10 +187,11 @@ def add_obo_nodes(self):
             oc = obo_client.OntologyClient(prefix=ns)
             for db_id, entry in oc.entries.items():
                 label = self.label(ns.upper(), db_id)
-                # There may be more general ways of doing this but this
-                # handles the special case of BFO nodes in the EFO
-                # ontology.
-                if ns == 'efo' and db_id.startswith('BFO'):
+                # Here we handle and isa relationships that point
+                # to an entry outside this ontology. The logic for recognizing
+                # these here is: if there is a : in the ID but the prefix is
+                # not for this namespace then we assume it's another namespace
+                if ':' in db_id and not db_id.startswith(ns.upper()):
                     label = db_id
                 nodes.append((label,
                               {'name': entry['name']}))
@@ -227,13 +228,15 @@ def add_obo_hierarchies(self):
                     mapped_rel = rel_mappings.get(rel)
                     if not mapped_rel:
                         continue
-                    source_label = self.label(ns.upper(), db_id)
-                    if ns == 'efo' and db_id.startswith('BFO'):
+                    if ':' in db_id and not db_id.startswith(ns.upper()):
                         source_label = db_id
+                    else:
+                        source_label = self.label(ns.upper(), db_id)
                     for target in targets:
-                        target_label = self.label(ns.upper(), target)
-                        if ns == 'efo' and target.startswith('BFO'):
+                        if ':' in target and not target.startswith(ns.upper()):
                             target_label = target
+                        else:
+                            target_label = self.label(ns.upper(), target)
                         if rel in reverse_rel:
                             av = (target_label,
                                   source_label,