def _decorate_existing_edges(self):
# This function decorates all existing edges in the knowledge graph with ICEES data, stored in EdgeAttributes
knowledge_graph = self.message.knowledge_graph
log = self.response
# Query ICEES for each edge in the knowledge graph that ICEES can provide data on (use known curies)
num_edges_obtained_icees_data_for = 0
edges_by_node_pair = self._get_edges_by_node_pair(
knowledge_graph) # Don't duplicate effort for parallel edges
for node_pair_key, node_pair_edges in edges_by_node_pair.items():
subject_key = node_pair_edges[0].subject
object_key = node_pair_edges[0].object
accepted_subject_synonyms = self._get_accepted_synonyms(
subject_key)
accepted_object_synonyms = self._get_accepted_synonyms(object_key)
if accepted_subject_synonyms and accepted_object_synonyms:
# Query ICEES for each possible combination of accepted subject/object synonyms
for subject_curie_to_try, object_curie_to_try in itertools.product(
accepted_subject_synonyms, accepted_object_synonyms):
qedge = QEdge(subject=subject_curie_to_try,
object=object_curie_to_try)
qedge.id = f"icees_e00"
log.debug(
f"Sending query to ICEES+ for {subject_curie_to_try}--{object_curie_to_try}"
)
p_value = self._get_icees_p_value_for_edge(qedge, log)
if p_value is not None:
num_edges_obtained_icees_data_for += len(
node_pair_edges)
new_edge_attribute = self._create_icees_edge_attribute(
p_value)
# Add the data as new EdgeAttributes on the existing edges with this subject/object ID
for edge in node_pair_edges:
if not edge.attributes:
edge.attributes = []
edge.attributes.append(new_edge_attribute)
# Don't worry about checking remaining synonym combos if we got results
break
if num_edges_obtained_icees_data_for:
log.info(
f"Overlayed {num_edges_obtained_icees_data_for} edges with exposures data from ICEES+"
)
else:
log.warning(
f"Could not find ICEES+ exposures data for any edges in the KG"
)
return self.response
def _add_virtual_edges(self, subject_qnode_key, object_qnode_key):
# This function adds ICEES exposures data as virtual edges between nodes with the specified qnode IDs
knowledge_graph = self.message.knowledge_graph
query_graph = self.message.query_graph
log = self.response
nodes_by_qg_id = self._get_nodes_by_qg_id(knowledge_graph)
subject_curies = set(nodes_by_qg_id.get(subject_qnode_key))
object_curies = set(nodes_by_qg_id.get(object_qnode_key))
# Determine which curies ICEES 'knows' about
known_subject_curies = {
curie
for curie in subject_curies if self._get_accepted_synonyms(curie)
}
known_object_curies = {
curie
for curie in object_curies if self._get_accepted_synonyms(curie)
}
num_node_pairs_recognized = 0
for subject_curie, object_curie in ou.get_node_pairs_to_overlay(
subject_qnode_key, object_qnode_key, query_graph,
knowledge_graph, log):
# Query ICEES only for synonyms it 'knows' about
if subject_curie in known_subject_curies and object_curie in known_object_curies:
accepted_subject_synonyms = self._get_accepted_synonyms(
subject_curie)
accepted_object_synonyms = self._get_accepted_synonyms(
object_curie)
for subject_synonym, object_synonym in itertools.product(
accepted_subject_synonyms, accepted_object_synonyms):
# qedge = QEdge(id=f"icees_{subject_synonym}--{object_synonym}",
# subject_key=subject_synonym,
# object_key=object_synonym)
qedge = QEdge(subject=subject_synonym,
object=object_synonym)
qedge.id = f"icees_{subject_synonym}--{object_synonym}"
log.debug(
f"Sending query to ICEES+ for {subject_synonym}--{object_synonym}"
)
p_value = self._get_icees_p_value_for_edge(qedge, log)
if p_value is not None:
num_node_pairs_recognized += 1
# Add a new virtual edge with this data
id, virtual_edge = self._create_icees_virtual_edge(
subject_curie, object_curie, p_value)
old_id = id
while id in knowledge_graph.edges:
id = old_id + f".{random.randint(10**(9-1), (10**9)-1)}"
knowledge_graph.edges[id] = virtual_edge
break # Don't worry about checking remaining synonym combos if we got results
# Add an 'empty' virtual edge (p-value of None) if we couldn't find any results for this node pair #1009
id, empty_virtual_edge = self._create_icees_virtual_edge(
subject_curie, object_curie, None)
while id in knowledge_graph.edges:
id = old_id + f".{random.randint(10**(9-1), (10**9)-1)}"
knowledge_graph.edges[id] = empty_virtual_edge
# Add a qedge to the query graph that corresponds to our new virtual edges
# new_qedge = QEdge(id=self.virtual_relation_label,
# subject_key=subject_qnode_key,
# object_key=object_qnode_key,
# type=self.icees_edge_type,
# option_group_id=ou.determine_virtual_qedge_option_group(subject_qnode_key, object_qnode_key,
# query_graph, log))
# Likely need to change this for TRAPI 1.0
new_qedge = QEdge(
subject=subject_qnode_key,
object=object_qnode_key,
predicate=self.icees_edge_type,
option_group_id=ou.determine_virtual_qedge_option_group(
subject_qnode_key, object_qnode_key, query_graph, log))
query_graph.edges[self.virtual_relation_label] = new_qedge
if num_node_pairs_recognized:
log.info(
f"ICEES+ returned data for {num_node_pairs_recognized} node pairs"
)
else:
log.warning(
f"Could not find ICEES+ exposures data for any {subject_qnode_key}--{object_qnode_key} node pairs"
)