def upton(iterable, truth_set, cutoff=1e-3, vmin=1/(3*60706.)):
def genchunks():
nsmall, ones = 0, 0
for i, chunk in enumerate(iterable):
if i % 100000 == 0:
if i > 0:
print i, chunk[0].chrom, chunk[0].start
if len(chunk) < 8:
nsmall += 1
continue
mafs = (float(x.mafs) for x in chunk)
score = sum((1.0 - m)**2.0 for m in mafs if m < cutoff) / float(len(chunk))
if score == 1:
ones += 1
continue
yield chunk, score
sys.stderr.write("%d (%.2f%%) removed for being too short\n" % (nsmall,
100.0 * nsmall / float(i)))
sys.stderr.write("%d (%.2f%%) removed for having score of 1\n" % (ones,
100.0 * ones /float(i)))
print >>sys.stderr, i, "total chunks"
# NOTE: this is for humvar only. not needed for clinvar.
def humvar_pathogenic(d):
return d['class'] == "deleterious"
if "humvar" in truth_set:
res = evaluate(genchunks(), truth_set, is_pathogenic=humvar_pathogenic)
else:
res = evaluate(genchunks(), truth_set)
print metrics(res[True], res[False], "upton.auc.png")
def rvis(truth_set):
interval = namedtuple('interval', ['chrom', 'start', 'end'])
def genregions():
for d in ts.reader("rvis.bed"):
score = float(d['pct'])
chunk = [interval(d['chrom'], int(d['start']), int(d['end']))]
yield chunk, -score
res = evaluate(genregions(), truth_set)
print metrics(res[True], res[False], "rvis.auc.png")