def db2ppi(list_sep):
"""
read dbfile and convert complexes to ppi for faster quering
"""
ppi_db = nx.Graph()
for members in list_sep:
for pairs in st.fast_comb(members.split("#"), 2):
ppi_db.add_edge(str.upper(pairs[0]), str.upper(pairs[1]))
ppi_db.remove_edges_from(nx.selfloop_edges(ppi_db, keys=True))
return ppi_db
def pairwise(self):
"""
performs pairwise comparison
"""
for pairs in st.fast_comb(np.array(self.members), 2):
self.calc_corr(pairs)
self.calc_diff(*pairs)
self.calc_shift([x.get_acc() for x in pairs])
# now need to average
self.cor = np.mean(self.cor, axis=0)
self.diff = np.mean(self.diff, axis=0)
self.shifts = np.mean(self.shifts)
def combine_all(G, gaf, t):
"""
permute all of blocks of whatever
"""
go_type = ["CC", "MF", "BP"]
out = []
for go in go_type:
k = [scr(G, gaf, x[0], x[1], go) for x in list(st.fast_comb(t, 2))]
out.append(st.mean(k))
# add to out the mean of the three Ontologies
# TODO check mean or sum
out.append(sum(out))
return out
def overlap_net(ppi_network, mb, over=0.5):
"""
calculates the overlap between a network and a list
"""
match, nomatch = 0, 0
mb = re.split(r"#", mb)
if len(mb) < 2:
return True
for pairs in st.fast_comb(mb, 2):
if ppi_network.has_edge(pairs[0], pairs[1]):
match += 1
else:
nomatch += 1
if match / (nomatch + match) >= over:
return True
else:
return False