def make_kg2_graph(input_file_name: str, test_mode: bool = False):
nodes = []
edges = []
df = pd.read_csv(input_file_name)
for idx in range(len(df)):
if not df['status'].isna()[idx]:
status = df['status'][idx].lower()
else:
status = "unknown_status"
if not df['phase'].isna()[idx]:
phase = df['phase'][idx].lower().replace(" ",
"_").replace("/", "_or_")
else:
phase = "unknown_phase"
relation = "clinically_tested_" + status + "_" + phase
edge_dict = kg2_util.make_edge(
subject_id=DRUGBANK_CURIE + df['drug_id'][idx],
object_id=UMLS_CURIE + df['ind_id'][idx],
relation=REPODB_IRI + '#' +
kg2_util.convert_snake_case_to_camel_case(relation),
relation_curie=REPODB_CURIE + relation,
predicate_label=relation,
provided_by=REPODB_IRI,
update_date=None)
if not df['NCT'].isna()[idx]:
edge_dict['publications'].append(NCT_CUTRIE + df['NCT'][idx])
edge_dict['publications info'][
NCT_CUTRIE +
df['NCT'][idx]] = CLINICALTRIALS_IRI + df['NCT'][idx]
edges.append(edge_dict)
return {'nodes': nodes, 'edges': edges}
def make_edge(subject_curie_id: str, object_curie_id: str,
predicate_label: str, update_date: str):
relation = kg2_util.BIOLINK_CATEGORY_BASE_IRI + kg2_util.convert_snake_case_to_camel_case(
predicate_label)
relation_curie = kg2_util.BIOLINK_CURIE_PREFIX + ':' + predicate_label.replace(
' ', '_')
rel = kg2_util.make_edge(subject_curie_id, object_curie_id, relation,
relation_curie, predicate_label,
UNIPROTKB_BASE_IRI, update_date)
return rel
def make_edge(subject_id: str,
object_id: str,
predicate_label: str,
update_date: str = None,
publications: list = None):
relation = CHEMBL_BASE_IRI_PREDICATE + kg2_util.convert_snake_case_to_camel_case(predicate_label)
if publications is None:
publications = []
return {'subject': subject_id,
'object': object_id,
'edge label': predicate_label,
'relation': relation,
'relation curie': 'CHEMBL:' + predicate_label,
'negated': False,
'publications': publications,
'publications info': {},
'update date': update_date,
'provided by': CHEMBL_KB_IRI}
def make_rel(preds_dict: dict, subject_curie: str, object_curie: str,
predicate: str, pmid: str, pub_date: str, sentence: str,
subject_score: str, object_score: str, negated: bool):
key = subject_curie + '-' + predicate + '-' + object_curie
key_val = preds_dict.get(key, None)
publication_curie = 'PMID:' + pmid
publication_info_dict = {
'publication date': pub_date,
'sentence': sentence,
'subject score': subject_score,
'object score': object_score
}
if key_val is None:
relation_type = predicate.lower()
if relation_type != 'xref':
relation_iri = kg2_util.convert_snake_case_to_camel_case(
relation_type.replace(' ', '_'))
relation_iri = relation_iri[0].lower() + relation_iri[1:]
relation_iri = SEMMEDDB_IRI + '#' + relation_iri
relation_curie = 'SEMMEDDB:' + relation_type
else:
relation_curie = 'OBO:xref'
relation_iri = prefixcommons.expand_uri(relation_curie)
edge_dict = kg2_util.make_edge(subject_curie, object_curie,
relation_iri, relation_curie,
relation_type, SEMMEDDB_IRI,
curr_timestamp)
edge_dict['publications'] = [publication_curie]
edge_dict['publications info'] = {
publication_curie: publication_info_dict
}
edge_dict['negated'] = negated
preds_dict[key] = edge_dict
else:
key_val['publications info'][publication_curie] = publication_info_dict
key_val['publications'] = key_val['publications'] + [publication_curie]
def make_kg2_graph(input_file_name: str, test_mode: bool = False):
nodes = []
edges = []
line_ctr = 0
update_date = None
with open(input_file_name, 'r') as input_file:
for line in input_file:
line = line.rstrip("\n")
if line.startswith('#'):
update_date = line.replace('#', '')
continue
if line.startswith('gene_name\t'):
continue
line_ctr += 1
if test_mode and line_ctr > 10000:
break
fields = line.split("\t")
[gene_name,
gene_claim_name,
entrez_id,
interaction_claim_source,
interaction_types,
drug_claim_name,
drug_claim_primary_name,
drug_name,
drug_chembl_id,
PMIDs] = fields
if entrez_id != "":
object_curie_id = 'NCBIGene:' + entrez_id
if drug_chembl_id != "":
subject_curie_id = 'CHEMBL.COMPOUND:' + drug_chembl_id
else:
if drug_claim_name != "":
node_pubs_list = []
subject_curie_id = None
if interaction_claim_source == "GuideToPharmacologyInteractions":
subject_curie_id = GTPI_CURIE_PREFIX + ':' + drug_claim_name
pmid_match = RE_PMID.match(drug_claim_primary_name)
if pmid_match is not None:
node_pubs_list = [pmid_match[2].replace(' ', '').strip()]
node_name = pmid_match[1].strip()
else:
node_name = drug_claim_primary_name
node_iri = GTPI_IRI_BASE + GTPI_LIGAND_SUFFIX + drug_claim_name
provided_by = GTPI_IRI_BASE
elif interaction_claim_source == "TTD":
subject_curie_id = TTD_CURIE_PREFIX + ':' + drug_claim_name
node_name = drug_claim_primary_name
node_iri = TTD_IRI_BASE + drug_claim_name
provided_by = TTD_IRI_BASE
if subject_curie_id is not None:
node_dict = kg2_util.make_node(subject_curie_id,
node_iri,
node_name,
'chemical_substance',
update_date,
provided_by)
node_dict['publications'] = node_pubs_list
nodes.append(node_dict)
if subject_curie_id is None:
print("DGIDB: no controlled ID was provided for this drug: " + drug_claim_primary_name + "; source DB: " + interaction_claim_source, file=sys.stderr)
continue
if interaction_types == "":
interaction_types = "affects"
pmids_list = []
if PMIDs.strip() != "":
pmids_list = [('PMID:' + pmid.strip()) for pmid in PMIDs.split(',')]
interaction_list = interaction_types.split(',')
for interaction in interaction_list:
interaction = interaction.replace(' ', '_')
edge_dict = kg2_util.make_edge(subject_curie_id,
object_curie_id,
DGIDB_BASE_IRI + '/' +
kg2_util.convert_snake_case_to_camel_case(interaction),
DGIDB_CURIE_PREFIX + ':' + interaction,
interaction,
DGIDB_BASE_IRI,
update_date)
edge_dict['publications'] = pmids_list
edges.append(edge_dict)
return {'nodes': nodes,
'edges': edges}
# Create the new edge(s) based on this SemMedDB row
for rel_to_make in get_rels_to_make_for_row(subject_cui_str,
object_cui_str, predicate):
subject_curie = rel_to_make[0]
object_curie = rel_to_make[1]
edge_label = rel_to_make[2]
# Exclude self-edges for certain types of predicates
if subject_curie != object_curie or edge_label.lower(
) not in EDGE_LABELS_EXCLUDE_FOR_LOOPS:
make_rel(edges_dict, subject_curie, object_curie, edge_label,
pmid, pub_date, sentence, subject_score, object_score,
negated)
if predicate not in nodes_dict:
relation_iri = kg2_util.convert_snake_case_to_camel_case(
predicate.lower().replace(' ', '_'))
relation_iri = SEMMEDDB_IRI + '#' + relation_iri
nodes_dict[predicate] = kg2_util.make_node(
id='SEMMEDDB:' + predicate.lower(),
iri=relation_iri,
name=predicate.lower(),
category_label="relationship type",
update_date=curr_timestamp,
provided_by=SEMMEDDB_IRI)
out_graph = {
'edges': [rel_dict for rel_dict in edges_dict.values()],
'nodes': [node_dict for node_dict in nodes_dict.values()]
}
for rel_dict in out_graph['edges']:
if len(rel_dict['publications']) > 1:
rel_dict['publications'] = list(set(rel_dict['publications']))