def open_ontology(path=None):
""" builds a networkx graph from obo parsed data
Returns:
networkx graph object, alt IDs and obsolete IDs
"""
header, entries = load_hpo_database(path)
graph = ICSimilarity()
# track alternate HPO IDs (since we use HPO IDs as node IDs)
alt_ids = {}
obsolete_ids = set()
# add the hpo header values as attributes for the graph
for header_id in header:
graph.graph[header_id] = header[header_id]
for entry in entries:
add_entry(graph, entry, alt_ids, obsolete_ids)
return graph, alt_ids, obsolete_ids
def setUp(self):
""" construct a ICSimilarity object for unit tests
"""
path = os.path.join(os.path.dirname(__file__), "data", "obo.txt")
ontology = Ontology(path)
graph = ontology.get_graph()
self.hpo_terms = {
"person_01": ["HP:0000924"],
"person_02": ["HP:0000118", "HP:0002011"],
"person_03": ["HP:0000707", "HP:0002011"]
}
self.hpo_graph = ICSimilarity(self.hpo_terms, graph)
def get_graph(self):
""" builds a networkx graph from obo parsed data
Returns:
networkx graph object
"""
graph = ICSimilarity()
# add the hpo header values as attributes for the graph
for header_id in self.hpo_header:
graph.graph[header_id] = self.hpo_header[header_id]
for entry in self.hpo_list:
tags = entry.tags
# ignore obsolete HPO entries
if self.is_obsolete(tags):
self.obsolete_ids.add(str(tags["id"][0]))
continue
node_id = str(tags["id"][0])
graph.add_node(node_id)
# make sure we can convert between HPO ID and their alternate IDs
self.track_alt_ids(tags, node_id)
# include the attribute data for the node
self.add_hpo_attributes_to_node(graph, node_id, tags)
# add the predecessors to the node
if "is_a" in tags:
for predecessor in tags["is_a"]:
predecessor = str(predecessor)
graph.add_edge(predecessor, node_id)
return graph
class TestICSimilarityPy(unittest.TestCase):
""" class to test ICSimilarity
"""
def setUp(self):
""" construct a ICSimilarity object for unit tests
"""
path = os.path.join(os.path.dirname(__file__), "data", "obo.txt")
ontology = Ontology(path)
graph = ontology.get_graph()
self.hpo_terms = {
"person_01": ["HP:0000924"],
"person_02": ["HP:0000118", "HP:0002011"],
"person_03": ["HP:0000707", "HP:0002011"]
}
self.hpo_graph = ICSimilarity(self.hpo_terms, graph)
def test_get_term_count(self):
""" check that get_term_count works correctly
All of the counts here are derived from their usage in self.hpo_terms
"""
# check that we count the term usage (and subterms correctly)
self.assertEqual(self.hpo_graph.get_term_count("HP:0000118"), 5)
self.assertEqual(self.hpo_graph.get_term_count("HP:0000707"), 3)
self.assertEqual(self.hpo_graph.get_term_count("HP:0002011"), 2)
# check that a terminal node, only used once in the probands, has a
# count of 1
self.assertEqual(self.hpo_graph.get_term_count("HP:0000924"), 1)
# check the term/subterm count for a term that isn't used within any of
# he probands, but which all of the used terms descend from.
self.assertEqual(self.hpo_graph.get_term_count("HP:0000001"), 5)
def test_calculate_information_content(self):
""" check that calculate_information_content works correctly
"""
# check that the top node has an information content of 0
self.assertEqual(self.hpo_graph.calculate_information_content("HP:0000001"), \
0)
# check the information content for a terminal node
self.assertAlmostEqual(self.hpo_graph.calculate_information_content("HP:0000924"), \
1.6094379)
# check the information content for a node that is somewhat distant, but
# which has some descendant nodes that need to be included in the term
# count
self.assertAlmostEqual(self.hpo_graph.calculate_information_content("HP:0000707"), \
0.5108256)
def test_get_most_informative_ic(self):
""" check that get_most_informative_ic works correctly
"""
# check the most informative information content for two nodes where
# every common ancestor is the ancestor of all terms used in the probands
self.assertAlmostEqual(self.hpo_graph.get_most_informative_ic("HP:0000707", \
"HP:0000924"), 0)
# check the most informative information content for two nodes where
# both nodes are somewhat down the HPO graph
self.assertAlmostEqual(self.hpo_graph.get_most_informative_ic("HP:0000707", \
"HP:0002011"), 0.5108256)
# check the most informative information content for two identical nodes
self.assertAlmostEqual(self.hpo_graph.get_most_informative_ic("HP:0000924", \
"HP:0000924"), 1.6094379)
