def _answer_query_using_bte(self, input_qnode_key: str, output_qnode_key: str, qg: QueryGraph,
answer_kg: QGOrganizedKnowledgeGraph, valid_bte_inputs_dict: Dict[str, Set[str]],
log: ARAXResponse) -> Tuple[QGOrganizedKnowledgeGraph, Set[str]]:
accepted_curies = set()
qedge_key = next(qedge_key for qedge_key in qg.edges)
qedge = qg.edges[qedge_key]
input_qnode = qg.nodes[input_qnode_key]
output_qnode = qg.nodes[output_qnode_key]
# Send this single-edge query to BTE, input curie by input curie (adding findings to our answer KG as we go)
for curie in input_qnode.id:
# Consider all different combinations of qnode types (can be multiple if gene/protein)
for input_qnode_category, output_qnode_category in itertools.product(input_qnode.category, output_qnode.category):
if eu.get_curie_prefix(curie) in valid_bte_inputs_dict['curie_prefixes']:
accepted_curies.add(curie)
try:
loop = asyncio.new_event_loop()
seqd = SingleEdgeQueryDispatcher(input_cls=input_qnode_category,
output_cls=output_qnode_category,
pred=qedge.predicate,
input_id=eu.get_curie_prefix(curie),
values=eu.get_curie_local_id(curie),
loop=loop)
log.debug(f"Sending query to BTE: {curie}-{qedge.predicate if qedge.predicate else ''}->{output_qnode_category}")
seqd.query()
reasoner_std_response = seqd.to_reasoner_std()
except Exception:
trace_back = traceback.format_exc()
error_type, error, _ = sys.exc_info()
log.error(f"Encountered a problem while using BioThings Explorer. {trace_back}",
error_code=error_type.__name__)
return answer_kg, accepted_curies
else:
answer_kg = self._add_answers_to_kg(answer_kg, reasoner_std_response, input_qnode_key, output_qnode_key, qedge_key, log)
return answer_kg, accepted_curies
def _get_best_equivalent_bte_curie(equivalent_curies: List[str], node_category: str) -> str:
# Curie prefixes in order of preference for different node types (not all-inclusive)
preferred_node_prefixes_dict = {'chemical_substance': ['CHEMBL.COMPOUND', 'CHEBI'],
'protein': ['UNIPROTKB', 'PR'],
'gene': ['NCBIGENE', 'ENSEMBL', 'HGNC', 'GO'],
'disease': ['DOID', 'MONDO', 'OMIM', 'MESH'],
'phenotypic_feature': ['HP', 'OMIM'],
'anatomical_entity': ['UBERON', 'FMA', 'CL'],
'pathway': ['REACTOME'],
'biological_process': ['GO'],
'cellular_component': ['GO']}
prefixes_in_order_of_preference = preferred_node_prefixes_dict.get(eu.convert_string_to_snake_case(node_category), [])
equivalent_curies.sort()
# Pick the curie that uses the (relatively) most preferred prefix
lowest_ranking = 10000
best_curie = None
for curie in equivalent_curies:
uppercase_prefix = eu.get_curie_prefix(curie).upper()
if uppercase_prefix in prefixes_in_order_of_preference:
ranking = prefixes_in_order_of_preference.index(uppercase_prefix)
if ranking < lowest_ranking:
lowest_ranking = ranking
best_curie = curie
# Otherwise, just try to pick one that isn't 'NAME:___'
if not best_curie:
non_name_curies = [curie for curie in equivalent_curies if eu.get_curie_prefix(curie).upper() != 'NAME']
best_curie = non_name_curies[0] if non_name_curies else equivalent_curies[0]
return best_curie