def expand_disont_disease(orangeboard, node):
disont_id = node.name
child_disease_ids_dict = QueryDisont.query_disont_to_child_disonts_desc(
disont_id)
for child_disease_id in child_disease_ids_dict.keys():
target_node = orangeboard.add_node(
'disont_disease',
child_disease_id,
desc=child_disease_ids_dict[child_disease_id])
orangeboard.add_rel('is_parent_of', 'DiseaseOntology', node,
target_node)
mesh_ids_set = QueryDisont.query_disont_to_mesh_id(disont_id)
for mesh_id in mesh_ids_set:
uniprot_ids_dict = QueryDisGeNet.query_mesh_id_to_uniprot_ids_desc(
mesh_id)
for uniprot_id in uniprot_ids_dict.keys():
source_node = orangeboard.add_node(
'uniprot_protein',
uniprot_id,
desc=uniprot_ids_dict[uniprot_id])
orangeboard.add_rel('gene_assoc_with', 'DisGeNet', source_node,
node)
## query for phenotypes associated with this disease
phenotype_id_dict = QueryBioLink.get_phenotypes_for_disease_desc(disont_id)
for phenotype_id_str in phenotype_id_dict.keys():
phenotype_node = orangeboard.add_node(
'phenont_phenotype',
phenotype_id_str,
desc=phenotype_id_dict[phenotype_id_str])
orangeboard.add_rel('phenotype_assoc_with', 'BioLink', phenotype_node,
node)
def expand_disease(self, node):
assert node.nodetype == "disease"
disease_name = node.name
gene_ontology_dict = QuerySciGraph.get_gene_ontology_curie_ids_for_disease_curie_id(disease_name)
for gene_ontology_curie_id_str, gene_ontology_term_dict in gene_ontology_dict.items():
gene_ontology_type_str = gene_ontology_term_dict["ontology"].replace(" ", "_")
target_node = self.add_node_smart(gene_ontology_type_str, gene_ontology_curie_id_str,
desc=gene_ontology_term_dict["name"])
if target_node is not None:
predicate_str = gene_ontology_term_dict["predicate"].replace(" ", "_")
self.orangeboard.add_rel("affects", "Monarch_SciGraph", node, target_node, extended_reltype=predicate_str)
if "OMIM:" in disease_name:
self.expand_genetic_condition(node)
return
if "MONDO:" in disease_name:
self.expand_mondo_disease(node)
return
# if we get here, this is a Disease Ontology disease
disont_id = disease_name
child_disease_ids_dict = QueryDisont.query_disont_to_child_disonts_desc(disont_id)
for child_disease_id in child_disease_ids_dict.keys():
target_node = self.add_node_smart('disease', child_disease_id,
desc=child_disease_ids_dict[child_disease_id])
if target_node is not None:
self.orangeboard.add_rel('subclass_of', 'DiseaseOntology',
target_node, node, extended_reltype="subclass_of")
mesh_ids_set = QueryDisont.query_disont_to_mesh_id(disont_id)
for mesh_id in mesh_ids_set:
uniprot_ids_dict = QueryDisGeNet.query_mesh_id_to_uniprot_ids_desc(mesh_id)
for uniprot_id in uniprot_ids_dict.keys():
assert '-' not in uniprot_id
source_node = self.add_node_smart('protein', uniprot_id,
desc=uniprot_ids_dict[uniprot_id])
if source_node is not None:
self.orangeboard.add_rel("gene_associated_with_condition", "DisGeNet", source_node,
node, extended_reltype="gene_associated_with_condition")
# query for phenotypes associated with this disease
phenotype_id_dict = QueryBioLink.get_phenotypes_for_disease_desc(disont_id)
for phenotype_id_str in phenotype_id_dict.keys():
phenotype_node = self.add_node_smart("phenotypic_feature", phenotype_id_str,
desc=phenotype_id_dict[phenotype_id_str])
if phenotype_node is not None:
self.orangeboard.add_rel("has_phenotype", 'BioLink', node, phenotype_node, extended_reltype="has_phenotype")
def test_query_disont_to_child_disonts_desc(self):
ret_dict = QD.query_disont_to_child_disonts_desc(
"DOID:9352") # type 2 diabetes mellitus
known_dict = {
'DOID:1837': 'diabetic ketoacidosis',
'DOID:10182': 'diabetic peripheral angiopathy',
'DOID:11712': 'lipoatrophic diabetes'
}
self.assertDictEqual(ret_dict, known_dict)
def get_mesh_term_for_all(curie_id, description):
"""
Takes a curie ID, detects the ontology from the curie id, and then finds the mesh term
Params:
curie_id - A string containing the curie id of the node. Formatted <source abbreviation>:<number> e.g. DOID:8398
description - A string containing the English name for the node
current functionality (+ means has it, - means does not have it)
"Reactome" +
"GO" - found gene conversion but no biological process conversion
"UniProt" +
"HP" - +
"UBERON" +
"CL" - not supposed to be here?
"NCBIGene" +
"DOID" +
"OMIM" +
"ChEMBL" +
"""
if type(description) != str:
description = str(description)
curie_list = curie_id.split(':')
names = None
if QueryNCBIeUtils.is_mesh_term(description):
return [description + '[MeSH Terms]']
names = NormGoogleDistance.get_mesh_from_oxo(curie_id)
if names is None:
if curie_list[0].lower().startswith("react"):
res = QueryNCBIeUtils.get_reactome_names(curie_list[1])
if res is not None:
names = res.split('|')
elif curie_list[0] == "GO":
pass
elif curie_list[0].startswith("UniProt"):
res = QueryNCBIeUtils.get_uniprot_names(curie_list[1])
if res is not None:
names = res.split('|')
elif curie_list[0] == "HP":
names = QueryNCBIeUtils.get_mesh_terms_for_hp_id(curie_id)
elif curie_list[0] == "UBERON":
if curie_id.endswith('PHENOTYPE'):
curie_id = curie_id[:-9]
mesh_id = QueryEBIOLS.get_mesh_id_for_uberon_id(curie_id)
names = []
for entry in mesh_id:
if len(entry.split('.')) > 1:
uids=QueryNCBIeUtils.get_mesh_uids_for_mesh_tree(entry.split(':')[1])
for uid in uids:
try:
uid_num = int(uid.split(':')[1][1:]) + 68000000
names += QueryNCBIeUtils.get_mesh_terms_for_mesh_uid(uid_num)
except IndexError:
uid_num = int(uid)
names += QueryNCBIeUtils.get_mesh_terms_for_mesh_uid(uid_num)
else:
try:
uid = entry.split(':')[1]
uid_num = int(uid[1:]) + 68000000
names += QueryNCBIeUtils.get_mesh_terms_for_mesh_uid(uid_num)
except IndexError:
uid_num = int(entry)
names += QueryNCBIeUtils.get_mesh_terms_for_mesh_uid(uid_num)
if len(names) == 0:
names = None
else:
names[0] = names[0] + '[MeSH Terms]'
elif curie_list[0] == "NCBIGene":
gene_id = curie_id.split(':')[1]
names = QueryNCBIeUtils.get_pubmed_from_ncbi_gene(gene_id)
elif curie_list[0] == "DOID":
mesh_id = QueryDisont.query_disont_to_mesh_id(curie_id)
names = []
for uid in mesh_id:
uid_num = int(uid[1:]) + 68000000
name = QueryNCBIeUtils.get_mesh_terms_for_mesh_uid(uid_num)
if name is not None:
names += name
if len(names) == 0:
names = None
else:
names[0] = names[0] + '[MeSH Terms]'
elif curie_list[0] == "OMIM":
names = QueryNCBIeUtils.get_mesh_terms_for_omim_id(curie_list[1])
elif curie_list[0] == "ChEMBL":
chembl_id = curie_id.replace(':', '').upper()
mesh_id = QueryMyChem.get_mesh_id(chembl_id)
if mesh_id is not None:
mesh_id = int(mesh_id[1:]) + 68000000
names = QueryNCBIeUtils.get_mesh_terms_for_mesh_uid(mesh_id)
if names is not None:
if type(names) == list:
for name in names:
if name.endswith('[MeSH Terms]'):
return [name]
return names
return [description.replace(';', '|')]
def seed_and_expand_kg_q2(num_expansions=3, seed_parts=None):
drug_dis_df = pandas.read_csv('../../data/q2/q2-drugandcondition-list.txt',
sep='\t')
if seed_parts is None or 'conditions' in seed_parts:
print('=====================> seeding disease nodes for Q2')
first_row = True
mesh_terms_set = set()
mesh_term_to_curie_ids_dict = dict()
curie_ids_for_df = []
for index, row in drug_dis_df.iterrows():
mesh_term = row['Condition']
if mesh_term not in mesh_terms_set:
mesh_term_to_curie_ids_dict[mesh_term] = None
mesh_terms_set.add(mesh_term)
curie_ids = get_curie_ont_ids_for_mesh_term(mesh_term)
if len(curie_ids) > 0:
assert type(curie_ids) == list
for curie_id in curie_ids:
if 'DOID:' in curie_id:
disont_desc = QueryDisont.query_disont_to_label(
curie_id)
ob.add_node('disont_disease',
curie_id,
desc=disont_desc,
seed_node_bool=first_row)
mesh_term_to_curie_ids_dict[mesh_term] = curie_id
first_row = False
else:
if 'HP:' in curie_id:
ob.add_node('phenont_phenotype',
curie_id,
desc=mesh_term,
seed_node_bool=first_row)
mesh_term_to_curie_ids_dict[
mesh_term] = curie_id
first_row = False
else:
assert False ## should never get here
curie_ids_for_df.append(mesh_term_to_curie_ids_dict[mesh_term])
drug_dis_df['CURIE_ID'] = pandas.Series(curie_ids_for_df,
index=drug_dis_df.index)
drug_dis_df.to_csv(
'../../data/q2/q2-drugandcondition-list-mapped-output.txt',
sep='\t')
## triple-expand the knowledge graph
for _ in range(0, num_expansions):
bne.expand_all_nodes()
if seed_parts is None or 'drugs' in seed_parts:
print('=====================> seeding drug nodes for Q2')
first_row = True
all_drugs = set()
for index, row in drug_dis_df.iterrows():
drug_name = row['Drug'].lower()
all_drugs.add(drug_name)
fda_drug_df = pandas.read_csv('../../data/q2/drugset2017_filt.txt',
sep='\t')
for index, row in fda_drug_df.iterrows():
drug_name = row['NAME'].lower()
all_drugs.add(drug_name)
for drug_name in all_drugs:
print(drug_name)
chembl_ids = QueryChEMBL.get_chembl_ids_for_drug(drug_name)
if chembl_ids is not None and len(chembl_ids) > 0:
chembl_id = next(iter(chembl_ids))
else:
chembl_id = ''
ob.add_node('pharos_drug',
drug_name,
desc=chembl_id,
seed_node_bool=first_row)
first_row = False
## triple-expand the knowledge graph
for _ in range(0, num_expansions):
bne.expand_all_nodes()
def setUpClass(cls):
cls.disont = QueryDisont()
def test_query_disont_to_child_disonts(self):
ret_set = QD.query_disont_to_child_disonts('DOID:9352')
known_set = {11712, 1837, 10182}
self.assertSetEqual(ret_set, known_set)
def test_query_disont_to_label(self):
ret_label = QD.query_disont_to_label("DOID:0050741")
self.assertEqual(ret_label, "alcohol dependence")