return cmp(x_num, y_num)
for ppi in ppis:
ppi.p1 = sorted(xrefs[ppi.p1].swissprot_id, cmp=s_cmp)[0]
ppi.p2 = sorted(xrefs[ppi.p2].swissprot_id, cmp=s_cmp)[0]
for ppi in unknown_ppis:
if ppi.p1 != '-':
ppi.p1 = sorted(xrefs[ppi.p1].swissprot_id, cmp=s_cmp)[0]
if ppi.p2 != '-':
ppi.p2 = sorted(xrefs[ppi.p2].swissprot_id, cmp=s_cmp)[0]
# Filter out small samples, self interactions and interactions with missing p2/p1
ppis = filter(lambda ppi: ppi.p1 != '-' and ppi.p2 != '-', ppis)
ppis = filter(lambda ppi: ppi.p1 != ppi.p2, ppis)
ppis = [ppi.split() for ppi in ppis]
ppis = set([ppi for sl in ppis for ppi in sl])
types = [ppi.get_reaction_type_string() for ppi in ppis]
type_counts = {k:0 for k in types}
for t in types:
type_counts[t] += 1
ppis = [ppi for ppi in ppis if type_counts[ppi.get_reaction_type_string()] >= 5]
fp1 = open('tmp/hprd_training_ppi.tsv', 'w')
fp2 = open('tmp/hprd_testing_ppi.tsv', 'w')
fp3 = open('tmp/unknown_hprd_ppi.tsv', 'w')
fp1.write("uniprot_a\tuniprot_a\tlabel\n")
fp2.write("uniprot_a\tuniprot_a\tlabel\n")
fp3.write("uniprot_a\tuniprot_a\tlabel\n")
