def curiefy_result(self, result):
"""
Convert subject, predicate and object IRIs to their respective CURIEs, where applicable
"""
if result['subject']['type'] == 'uri':
subject_curie = make_curie(result['subject']['value'])
if subject_curie != result['subject']['value']:
result['subject']['value'] = subject_curie
result['subject']['type'] = 'curie'
else:
logging.warning("Could not CURIEfy {}".format(
result['subject']['value']))
if result['object']['type'] == 'curie':
object_curie = make_curie(result['object']['value'])
if object_curie != result['object']['value']:
result['object']['value'] = object_curie
result['object']['type'] = 'curie'
else:
logging.warning("Could not CURIEfy {}".format(
result['object']['value']))
predicate_curie = make_curie(result['predicate']['value'])
if predicate_curie != result['predicate']['value']:
result['predicate']['value'] = predicate_curie
result['predicate']['type'] = 'curie'
else:
result['predicate']['type'] = 'uri'
return result
def add_edge_attribute(self, subject_iri: Union[URIRef, str],
object_iri: URIRef, predicate_iri: URIRef, key: str,
value: str) -> None:
"""
Adds an attribute to an edge, while taking into account whether the attribute
should be multi-valued.
Multi-valued properties will not contain duplicates.
The key may be a rdflib.URIRef or a URI string that maps onto a property name
as defined in `rdf_utils.property_mapping`.
If the nodes in the edge does not exist then they will be created
using subject_iri and object_iri.
If the edge itself does not exist then it will be created using
subject_iri, object_iri and predicate_iri.
Parameters
----------
subject_iri: [rdflib.URIRef, str]
The IRI of the subject node of an edge in rdflib.Graph
object_iri: rdflib.URIRef
The IRI of the object node of an edge in rdflib.Graph
predicate_iri: rdflib.URIRef
The IRI of the predicate representing an edge in rdflib.Graph
key: str
The name of the attribute. Can be a rdflib.URIRef or URI string
value: str
The value of the attribute
"""
if key.lower() in is_property_multivalued:
key = key.lower()
else:
if not isinstance(key, URIRef):
key = URIRef(key)
key = property_mapping.get(key)
if key is not None:
subject_curie = make_curie(subject_iri)
object_curie = make_curie(object_iri)
edge_label = process_iri(predicate_iri)
if is_curie(edge_label):
edge_label = curie_lookup(edge_label)
edge_key = generate_edge_key(subject_curie, edge_label,
object_curie)
attr_dict = self.graph.get_edge_data(subject_curie,
object_curie,
key=edge_key)
self._add_attribute(attr_dict, key, value)
def add_node(self, iri: URIRef) -> str:
"""
This method should be used by all derived classes when adding a node to
the networkx.MultiDiGraph. This ensures that a node's identifier is a CURIE,
and that it's 'iri' property is set.
Returns the CURIE identifier for the node in the networkx.MultiDiGraph
Parameters
----------
iri : rdflib.URIRef
IRI of a node
Returns
-------
str
The CURIE identifier of a node
"""
kwargs = {
'iri': str(iri),
}
if 'provided_by' in self.graph_metadata:
kwargs['provided_by'] = self.graph_metadata['provided_by']
n = make_curie(iri)
if n not in self.graph:
self.graph.add_node(n, **kwargs)
return n
def add_edge(self, subject_iri: URIRef, object_iri: URIRef, predicate_iri: URIRef) -> Tuple[str, str, str]:
"""
This method should be used by all derived classes when adding an edge to the networkx.MultiDiGraph.
This ensures that the `subject` and `object` identifiers are CURIEs, and that `edge_label` is in the correct form.
Returns the CURIE identifiers used for the `subject` and `object` in the
networkx.MultiDiGraph, and the processed `edge_label`.
Parameters
----------
subject_iri: rdflib.URIRef
Subject IRI for the subject in a triple
object_iri: rdflib.URIRef
Object IRI for the object in a triple
predicate_iri: rdflib.URIRef
Predicate IRI for the predicate in a triple
Returns
-------
Tuple[str, str, str]
A 3-nary tuple (of the form subject, object, predicate) that represents the edge
"""
s = self.add_node(subject_iri)
o = self.add_node(object_iri)
relation = make_curie(predicate_iri)
edge_label = process_iri(predicate_iri)
if ' ' in edge_label:
logging.debug("predicate IRI '{}' yields edge_label '{}' that not in snake_case form; replacing ' ' with '_'".format(predicate_iri, edge_label))
if edge_label.startswith(self.BIOLINK):
logging.debug("predicate IRI '{}' yields edge_label '{}' that starts with '{}'; removing IRI prefix".format(predicate_iri, edge_label, self.BIOLINK))
edge_label = edge_label.replace(self.BIOLINK, '')
if PrefixManager.is_curie(edge_label):
name = curie_lookup(edge_label)
if name:
logging.debug("predicate IRI '{}' yields edge_label '{}' that is actually a CURIE; Using its mapping instead: {}".format(predicate_iri, edge_label, name))
edge_label = name
else:
logging.debug("predicate IRI '{}' yields edge_label '{}' that is actually a CURIE; defaulting back to {}".format(predicate_iri, edge_label, self.DEFAULT_EDGE_LABEL))
edge_label = self.DEFAULT_EDGE_LABEL
kwargs = {
'subject': s,
'predicate': predicate_iri,
'object': o,
'relation': relation,
'edge_label': edge_label
}
if 'provided_by' in self.graph_metadata:
kwargs['provided_by'] = self.graph_metadata['provided_by']
key = generate_edge_key(s, edge_label, o)
if not self.graph.has_edge(s, o, key=key):
self.graph.add_edge(s, o, key=key, **kwargs)
return s, o, edge_label
def _add_attribute(self, attr_dict: Dict, key: str, value: str) -> None:
"""
Adds an attribute to the attribute dictionary, respecting whether or not
that attribute should be multi-valued.
Multi-valued attributes will not contain duplicates.
Some attributes are singular form of others. In such cases overflowing values
will be placed into the correlating multi-valued attribute.
For example, 'name' attribute will hold only one value while any additional
value will be stored as 'synonym' attribute.
Parameters
----------
attr_dict: dict
Dictionary representing the attribute set of a node or an edge in a networkx graph
key: str
The name of the attribute
value: str
The value of the attribute
"""
if key is None or key not in is_property_multivalued:
logging.warning(
"Discarding key {} as it is not a valid property.".format(key))
return
value = make_curie(process_iri(value))
if is_property_multivalued[key]:
if key not in attr_dict:
attr_dict[key] = [value]
elif value not in attr_dict[key]:
attr_dict[key].append(value)
else:
if key not in attr_dict:
attr_dict[key] = value
elif key == 'name':
self._add_attribute(attr_dict, 'synonym', value)
def add_edge(self, subject_iri: URIRef, object_iri: URIRef,
predicate_iri: URIRef) -> Tuple[str, str, str]:
"""
This method should be used by all derived classes when adding an edge to the networkx.MultiDiGraph.
This ensures that the subject and object identifiers are CURIEs, and that edge_label is in the correct form.
Returns the CURIE identifiers used for the subject and object in the
networkx.MultiDiGraph, and the processed edge_label.
Parameters
----------
subject_iri: rdflib.URIRef
Subject IRI for the subject in a triple
object_iri: rdflib.URIRef
Object IRI for the object in a triple
predicate_iri: rdflib.URIRef
Predicate IRI for the predicate in a triple
Returns
-------
Tuple[str, str, str]
A 3-nary tuple (of the form subject, object, predicate) that represents the edge
"""
s = self.add_node(subject_iri)
o = self.add_node(object_iri)
relation = make_curie(predicate_iri)
edge_label = process_iri(predicate_iri)
if ' ' in edge_label:
logging.debug(
"predicate IRI '{}' yields edge_label '{}' that not in snake_case form; replacing ' ' with '_'"
.format(predicate_iri, edge_label))
# TODO: shouldn't this move to the utilities function process_uri()
if edge_label.startswith(self.BIOLINK):
logging.debug(
"predicate IRI '{}' yields edge_label '{}' that starts with '{}'; removing IRI prefix"
.format(predicate_iri, edge_label, self.BIOLINK))
edge_label = edge_label.replace(self.BIOLINK, '')
# TODO: is there no way to get label of a CURIE?
# TODO: this should also move to the utilities function
# Any service? or preload required ontologies by prefix?
if ':' in edge_label:
logging.debug(
"edge label '{}' is a CURIE; defaulting back to 'related_to'".
format(edge_label))
logging.debug(
"predicate IRI '{}' yields edge_label '{}' that is actually a CURIE; defaulting back to {}"
.format(predicate_iri, edge_label, self.DEFAULT_EDGE_LABEL))
edge_label = self.DEFAULT_EDGE_LABEL
kwargs = {'relation': relation, 'edge_label': edge_label}
if 'provided_by' in self.graph_metadata:
kwargs['provided_by'] = self.graph_metadata['provided_by']
if self.graph.has_edge(s, o, key=edge_label):
logging.debug("{} -- {} --> {} edge already exists".format(
s, edge_label, o))
else:
self.graph.add_edge(s, o, key=edge_label, **kwargs)
return s, o, edge_label
"shortened life span",
"http://purl.obolibrary.org/obo/CHEBI_23367": "molecular entity",
"http://purl.obolibrary.org/obo/CHEBI_23888": "drug",
"http://purl.obolibrary.org/obo/CHEBI_51086": "chemical role",
"http://purl.obolibrary.org/obo/UPHENO_0001001": "phenotypic feature",
"http://purl.obolibrary.org/obo/GO_0008150": "biological_process",
"http://purl.obolibrary.org/obo/GO_0005575": "cellular component",
"http://purl.obolibrary.org/obo/SO_0000704": "gene",
"http://purl.obolibrary.org/obo/SO_0000110": "sequence feature",
"http://purl.obolibrary.org/obo/GENO_0000536": "genotype",
}
mapping = defaultdict(set)
for key, value in iri_mapping.items():
mapping[make_curie(key)].add(value)
for key, value in toolkit.generator.mappings.items():
mapping[key].update(value)
def walk(node, next_node_generator):
to_visit = {node: 0} # Dict[URIRef, Integer]
visited = {} # Dict[URIRef, Integer]
while to_visit != {}:
m, score = to_visit.popitem()
visited[m] = score
for t in next_node_generator(m):
if isinstance(t, tuple) and len(t) > 1:
n, s = t
def load_edges(self,
association='bl:ChemicalToGeneAssociation',
limit=None):
sparql = SPARQLWrapper(self.url)
query = render(self.count_query, {'association': association})
logging.debug(query)
sparql.setQuery(query)
sparql.setReturnFormat(JSON)
results = sparql.query().convert()
count = int(results['results']['bindings'][0]['triples']['value'])
logging.info("Expected triples for query: {}".format(count))
step = 1000
start = 0
for i in range(step, count + step, step):
end = i
query = render(self.edge_query, {
'association': association,
'offset': start,
'limit': step
})
sparql.setQuery(query)
results = sparql.query().convert()
node_list = set()
for r in results['results']['bindings']:
node_list.add("<{}>".format(r['subject']['value']))
node_list.add("<{}>".format(r['object']['value']))
start = end
self.load_nodes(node_list)
logging.info("Fetching edges...")
map = {}
for r in results['results']['bindings']:
s = r['subject']['value']
p = r['predicate']['value']
o = r['object']['value']
self.add_edge(s, o, p)
continue
# make_curie_result(r)
key = ((r['subject']['value'], r['object']['value']),
r['predicate']['value'])
if key in map:
# seen this triple before. look at properties
edge_property_key = r['edge_property_key']
edge_property_key_curie = make_curie(
edge_property_key['value'])
if edge_property_key_curie.startswith('bl:'):
edge_property_key_curie = edge_property_key_curie.split(
':')[1]
edge_property_value = r['edge_property_value']
if edge_property_value['type'] == 'uri':
edge_property_value_curie = make_curie(
edge_property_value['value'])
else:
edge_property_value_curie = edge_property_value[
'value']
map[key][
edge_property_key_curie] = edge_property_value_curie
else:
map[key] = {}
edge_property_key = r['edge_property_key']
edge_property_key_curie = make_curie(
edge_property_key['value'])
if edge_property_key_curie.startswith('bl:'):
edge_property_key_curie = edge_property_key_curie.split(
':')[1]
edge_property_value = r['edge_property_value']
if edge_property_value['type'] == 'uri':
edge_property_value_curie = make_curie(
edge_property_value['value'])
else:
edge_property_value_curie = edge_property_value[
'value']
map[key][
edge_property_key_curie] = edge_property_value_curie
logging.info("Loading edges...")
for key, properties in map.items():
self.graph.add_node(key[0][0])
self.graph.add_node(key[0][1])
if 'is_defined_by' not in properties and self.IS_DEFINED_BY:
properties['is_defined_by'] = self.IS_DEFINED_BY
if key[1].startswith('bl:'):
relation = key[1].split(':')[1]
else:
relation = key[1]
properties['edge_label'] = relation
if 'relation' not in properties:
properties['relation'] = relation
self.graph.add_edge(key[0][0], key[0][1], **properties)
map.clear()
if limit is not None and i > limit:
break
self.set_categories()
