def main(args):
import argparse
import gc
import pathlib
from sklearn.externals import joblib
import numpy as np
from peakachu import scoreUtils, utils
np.seterr(divide='ignore', invalid='ignore')
pathlib.Path(args.output).mkdir(parents=True, exist_ok=True)
model = joblib.load(args.model)
# more robust to check if a file is .hic
hic_info = utils.read_hic_header(args.path)
if hic_info is None:
hic = False
import cooler
Lib = cooler.Cooler(args.path)
chromosomes = Lib.chromnames[:]
#nam = args.path.split('.cool')[0]
else:
hic = True
chromosomes = utils.get_hic_chromosomes(args.path, args.resolution)
#nam = args.path.split('.hic')[0]
#nam = nam.split('/')[-1]
for key in chromosomes:
if key.startswith('chr'):
cname = key
else:
cname = 'chr'+key
if not hic:
X = scoreUtils.Chromosome(Lib.matrix(balance=args.balance, sparse=True).fetch(key).tocsr(),
model=model,
cname=cname, lower=args.lower,
upper=args.upper, res=args.resolution,
width=args.width)
else:
if args.balance:
X = scoreUtils.Chromosome(utils.csr_contact_matrix('KR', args.path, key, key, 'BP', args.resolution),
model=model,
cname=cname, lower=args.lower,
upper=args.upper, res=args.resolution,
width=args.width)
else:
X = scoreUtils.Chromosome(utils.csr_contact_matrix('NONE', args.path, key, key, 'BP', args.resolution),
model=model,
cname=cname, lower=args.lower,
upper=args.upper, res=args.resolution,
width=args.width)
result, R = X.score()
X.writeBed(args.output, result, R)
def main(args):
import argparse, gc, pathlib, os
import numpy as np
from sklearn.externals import joblib
from peakachu import scoreUtils, utils
np.seterr(divide='ignore', invalid='ignore')
pathlib.Path(args.output).mkdir(parents=True, exist_ok=True)
model = joblib.load(args.model)
hic_info = utils.read_hic_header(
args.path) # more robust to check if a file is .hic
if hic_info is None:
hic = False
else:
hic = True
if not hic:
import cooler
Lib = cooler.Cooler(args.path)
chromosomes = Lib.chromnames[:]
else:
chromosomes = list(hic_info['chromsizes'])
pre = utils.find_chrom_pre(chromosomes)
tmp = os.path.split(args.model)[1] # support full path
ccname = pre + tmp.split('.pk')[0].lstrip(
'chr') # ccname is consistent with chromosome labels in .hic / .cool
cikada = 'chr' + ccname.lstrip('chr') # cikada always has prefix "chr"
if not hic:
X = scoreUtils.Chromosome(Lib.matrix(
balance=args.balance, sparse=True).fetch(ccname).tocsr(),
model=model,
cname=cikada,
lower=args.lower,
upper=args.upper,
res=args.resolution,
width=args.width)
else:
if args.balance:
X = scoreUtils.Chromosome(utils.csr_contact_matrix(
'KR', args.path, ccname, ccname, 'BP', args.resolution),
model=model,
cname=cikada,
lower=args.lower,
upper=args.upper,
res=args.resolution,
width=args.width)
else:
X = scoreUtils.Chromosome(utils.csr_contact_matrix(
'NONE', args.path, ccname, ccname, 'BP', args.resolution),
model=model,
cname=cikada,
lower=args.lower,
upper=args.upper,
res=args.resolution,
width=args.width)
result, R = X.score()
X.writeBed(args.output, result, R)
def main(args):
from sklearn.externals import joblib
import gc
import pathlib
import straw
import numpy as np
from peakachu import trainUtils, utils
np.seterr(divide='ignore', invalid='ignore')
pathlib.Path(args.output).mkdir(parents=True, exist_ok=True)
# more robust to check if a file is .hic
hic_info = utils.read_hic_header(args.path)
if hic_info is None:
hic = False
else:
hic = True
coords = trainUtils.parsebed(args.bedpe, lower=2, res=args.resolution)
kde, lower, long_start, long_end = trainUtils.learn_distri_kde(coords)
if not hic:
import cooler
Lib = cooler.Cooler(args.path)
chromosomes = Lib.chromnames[:]
else:
chromosomes = utils.get_hic_chromosomes(args.path, args.resolution)
# train model per chromosome
positive_class = {}
negative_class = {}
for key in chromosomes:
if key.startswith('chr'):
chromname = key
else:
chromname = 'chr' + key
print('collecting from {}'.format(key))
if not hic:
X = Lib.matrix(balance=args.balance,
sparse=True).fetch(key).tocsr()
else:
if args.balance:
X = utils.csr_contact_matrix('KR', args.path, key, key, 'BP',
args.resolution)
else:
X = utils.csr_contact_matrix('NONE', args.path, key, key, 'BP',
args.resolution)
clist = coords[chromname]
try:
positive_class[chromname] = np.vstack(
(f
for f in trainUtils.buildmatrix(X, clist, width=args.width)))
neg_coords = trainUtils.negative_generating(
X, kde, clist, lower, long_start, long_end)
stop = len(clist)
negative_class[chromname] = np.vstack(
(f for f in trainUtils.buildmatrix(
X, neg_coords, width=args.width, positive=False, stop=stop)
))
except:
print(chromname, ' failed to gather fts')
for key in chromosomes:
if key.startswith('chr'):
chromname = key
else:
chromname = 'chr' + key
Xtrain = np.vstack(
(v for k, v in positive_class.items() if k != chromname))
Xfake = np.vstack(
(v for k, v in negative_class.items() if k != chromname))
print(chromname, 'pos/neg: ', Xtrain.shape[0], Xfake.shape[0])
model = trainUtils.trainRF(Xtrain, Xfake)
joblib.dump(model,
args.output + '/' + chromname + '.pkl',
compress=('xz', 3))