def decode(pwmid,
sample,
cutk=0,
pwmbase='transfac',
pos_threshold=np.log10(99),
chipseq=False):
import centipede
import millipede
import centipede_pbm as pbmcentipede
if sample in [None, 'Gm12878', 'Gm12878All']:
sequence = loadutils.Sequence(sample)
else:
indiv_idx = loadutils.read_individuals()
if pwmbase == 'transfac':
pwms = loadutils.transfac_pwms()
elif pwmbase == 'selex':
pwms = loadutils.selex_pwms()
motif = [
val['motif'] for val in pwms.itervalues() if val['AC'] == pwmid
][0]
sequence = loadutils.Sequence(sample, sample_idx=indiv_idx[sample])
if cutk != 0:
sequence.set_cutrate(sample=sample, k=cutk)
# use output from Centipede run
# 0 = Py code, 1 = R code
if sample is None:
handle = open(
"%s/cache/combined/pbmcentipede_short_%s.pkl" % (projpath, pwmid),
'r')
else:
handle = open(
"%s/cache/separate/pbmcentipede_short_%s_%s.pkl" %
(projpath, pwmid, sample), 'r')
output = cPickle.load(handle)
handle.close()
if cutk == 0:
idx = 0
elif cutk == 2:
idx = 1
elif cutk == 4:
idx = 2
footprint = output['footprint'][idx]
negbinparams = output['negbin'][idx]
prior = output['prior'][idx][0]
dhsprior = output['prior'][idx][1]
if sample in ['Gm12878', 'Gm12878All']:
location_file = "%s/cache/%s_locationsGm12878_Q%.1f.txt.gz" % (
projpath, pwmid, dhs)
else:
location_file = "%s/cache/%s_locations_Q%.1f.txt.gz" % (projpath,
pwmid, dhs)
# check file size
pipe = subprocess.Popen("zcat %s | wc -l" % location_file,
stdout=subprocess.PIPE,
shell=True)
Ns = int(pipe.communicate()[0].strip())
try:
chipobj = loadutils.ChipSeq('Gm12878', loadutils.factormap[pwmid])
except:
pass
readobj = loadutils.Dnase(sample=sample)
readhandle = loadutils.ZipFile(location_file)
loops = Ns / batch
if sample is None:
handle = gzip.open(
"%s/cache/combined/%s_short_bound.bed.gz" % (projpath, pwmid),
'wb')
else:
handle = gzip.open(
"%s/cache/separate/%s_%d_%s_short_bound.bed.gz" %
(projpath, pwmid, cutk, sample), 'wb')
towrite = [
'Chr', 'Start', 'Stop', 'Strand', 'PwmScore', 'LogPosOdds',
'LogPriorOdds', 'MultLikeRatio', 'NegBinLikeRatio', 'ChipseqReads'
]
handle.write('\t'.join(towrite) + '\n')
totalreads = []
for n in xrange(loops):
starttime = time.time()
# read locations from file
locations = readhandle.read(chunk=batch)
if sample not in [None, 'Gm12878', 'Gm12878All']:
# compute scores at locations for specific sample
locations = sequence.get_scores(locations, motif)
locations = sequence.filter_mappability(locations,
width=max([200, L / 2]))
# read in Dnase read data for locations
dnasereads, locations, subscores = readobj.getreads(locations,
width=max(
[200, L / 2]))
subscores = np.array(subscores).astype('float')
subscores = subscores.reshape(subscores.size, 1)
dnasetotal = dnasereads.sum(1)
print len(locations)
if chipseq:
chipreads = chipobj.getreads(locations, width=max([200, L / 2]))
else:
chipreads = None
# set null footprint distribution
if cutk == 0:
null = np.ones((1, L), dtype=float) / L
else:
null = sequence.getnull(locations, width=L / 2)
if L < 400:
dnasereads = np.hstack((dnasereads[:, 100 - L / 4:100 + L / 4],
dnasereads[:, 300 - L / 4:300 + L / 4]))
# if cutk==0:
logodds = centipede.decode(dnasereads,
dnasetotal,
null,
subscores,
footprint,
negbinparams[0],
negbinparams[1],
prior,
dhsprior,
chipreads=chipreads,
damp=damp)
# elif cutk==2:
# posterior = pbmcentipede.decode(reads, chipreads, subscores, footprint[1:], negbinparams[0], negbinparams[1], prior)
if not chipseq:
try:
chipreads = chipobj.get_total_reads(locations, width=400)
ignore = [loc.extend(['%.3f'%pos[0],'%.3f'%pos[1],'%.3f'%pos[2],'%.3f'%pos[3],'%d'%c]) \
for loc,pos,c in zip(locations,logodds,chipreads)]
except NameError:
ignore = [loc.extend(['%.3f'%pos[0],'%.3f'%pos[1],'%.3f'%pos[2],'%.3f'%pos[3]]) \
for loc,pos in zip(locations,logodds)]
locations = [loc for loc in locations if len(loc) > 5]
ignore = [handle.write('\t'.join(elem) + '\n') for elem in locations]
print time.time() - starttime
remain = Ns - loops * batch
locations = readhandle.read(chunk=remain)
if sample not in [None, 'Gm12878', 'Gm12878All']:
# compute scores at locations for specific sample
locations = sequence.get_scores(locations, motif)
locations = sequence.filter_mappability(locations, width=max([200, L / 2]))
dnasereads, locations, subscores = readobj.getreads(locations,
width=max([200,
L / 2]))
subscores = np.array(subscores)
subscores = subscores.reshape(subscores.size, 1)
dnasetotal = dnasereads.sum(1)
if chipseq:
chipreads = chipobj.get_total_reads(locations, width=200)
else:
chipreads = None
# set null footprint distribution
if cutk == 0:
null = np.ones((1, L), dtype=float) / L
else:
null = sequence.getnull(locations, width=L / 2)
if L < 400:
dnasereads = np.hstack((dnasereads[:, 100 - L / 4:100 + L / 4],
dnasereads[:, 300 - L / 4:300 + L / 4]))
logodds = centipede.decode(dnasereads,
dnasetotal,
null,
subscores,
footprint,
negbinparams[0],
negbinparams[1],
prior,
dhsprior,
chipreads=chipreads,
damp=damp)
if not chipseq:
try:
chipreads = chipobj.get_total_reads(locations, width=400)
ignore = [loc.extend(['%.3f'%pos[0],'%.3f'%pos[1],'%.3f'%pos[2],'%.3f'%pos[3],'%d'%c]) \
for loc,pos,c in zip(locations,logodds,chipreads)]
except NameError:
ignore = [loc.extend(['%.3f'%pos[0],'%.3f'%pos[1],'%.3f'%pos[2],'%.3f'%pos[3]]) \
for loc,pos in zip(locations,logodds)]
locations = [loc for loc in locations if len(loc) > 5]
ignore = [handle.write('\t'.join(elem) + '\n') for elem in locations]
readobj.close()
chipobj.close()
readhandle.close()
handle.close()
sequence.close()
def infer(pwmid, sample, pwm_thresh=8, pwmbase='transfac', chipseq=False):
import centipede_pbm as centipede
model = 'modelC'
if pwmbase == 'transfac':
pwms = loadutils.transfac_pwms()
elif pwmbase == 'selex':
pwms = loadutils.selex_pwms()
motif = [val['motif'] for val in pwms.itervalues()
if val['AC'] == pwmid][0]
if sample in [None, 'Gm12878', 'Gm12878All']:
sequence = loadutils.Sequence(sample)
else:
indiv_idx = loadutils.read_individuals()
sequence = loadutils.Sequence(sample, sample_idx=indiv_idx[sample])
if sample in ['Gm12878', 'Gm12878All']:
location_file = "/mnt/lustre/home/anilraj/histmod/cache/%s_locationsGm12878_Q%.1f.txt.gz" % (
pwmid, dhs)
else:
location_file = "/mnt/lustre/home/anilraj/histmod/cache/%s_locations_Q%.1f.txt.gz" % (
pwmid, dhs)
# check file size
pipe = subprocess.Popen("zcat %s | wc -l" % location_file,
stdout=subprocess.PIPE,
shell=True)
Ns = int(pipe.communicate()[0].strip())
# load scores
alllocations = []
pwm_cutoff = pwm_thresh + 1
while len(alllocations) < 100:
pwm_cutoff = pwm_cutoff - 1
handle = loadutils.ZipFile(location_file)
alllocations = handle.read(threshold=pwm_cutoff)
handle.close()
print "PWM Cutoff = %d" % pwm_cutoff
# subsample locations, if too many
if len(alllocations) > 100000:
scores = np.array([loc[-1] for loc in alllocations]).astype(float)
indices = np.argsort(scores)[-100000:]
alllocations = [alllocations[index] for index in indices]
print "Num of sites for learning, with pwm threshold of %d for %s = %d" % (
pwm_thresh, pwmid, len(alllocations))
if sample in [None, 'Gm12878', 'Gm12878All']:
locs_tolearn = alllocations
else:
# compute scores for specific sample at these locations
starttime = time.time()
locs_tolearn = sequence.get_scores(alllocations, motif)
print len(locs_tolearn), time.time() - starttime
# filter mappability
print "filtering out unmappable sites ..."
locs_tolearn = sequence.filter_mappability(locs_tolearn,
width=max([200, L / 2]))
# load reads and locations
print "loading dnase reads ..."
readobj = loadutils.Dnase(sample=sample)
dnasereads, locs_tolearn, subscores = readobj.getreads(
locs_tolearn, remove_outliers=True, width=max([200, L / 2]))
subscores = np.array(subscores)
subscores = subscores.reshape(subscores.size, 1)
dnasetotal = dnasereads.sum(1)
print "Num of mappable sites for learning for %s = %d" % (
pwmid, len(locs_tolearn))
if chipseq:
chipobj = loadutils.ChipSeq('Gm12878', loadutils.factormap[pwmid])
chipreads = chipobj.get_total_reads(locs_tolearn, width=200)
chipobj.close()
else:
chipreads = None
if L < 400:
dnasereads = np.hstack((dnasereads[:, 100 - L / 4:100 + L / 4],
dnasereads[:, 300 - L / 4:300 + L / 4]))
locs_tolearn = [list(loc) for loc in locs_tolearn]
footprints = []
priors = []
negbins = []
posteriors = []
null = np.ones((1, L), dtype=float) * 1. / L
posterior, footprint, negbinparams, prior = centipede.EM(dnasereads,
dnasetotal,
subscores,
null,
model=model,
restarts=2)
posteriors.append(posterior)
footprints.append(footprint)
negbins.append(negbinparams)
priors.append(prior)
chipobj = loadutils.ChipSeq('Gm12878', loadutils.factormap[pwmid])
controlobj = loadutils.ChipSeq('Gm12878', loadutils.controlmap[pwmid])
chipreads = chipobj.get_total_reads(locs_tolearn, width=400)
controlreads = controlobj.get_total_reads(locs_tolearn, width=200)
chipobj.close()
controlobj.close()
for posterior in posteriors:
logodds = np.log(posterior[:, 1] / posterior[:, 0])
logodds[logodds == np.inf] = logodds[logodds != np.inf].max()
logodds[logodds == -np.inf] = logodds[logodds != -np.inf].min()
R = stats.pearsonr(logodds, np.sqrt(chipreads))
R2 = stats.pearsonr(np.sqrt(dnasetotal), np.sqrt(chipreads))
handle = open(
'/mnt/lustre/home/anilraj/histmod/cache/chipseq_peaks/%s_peaks.bed'
% loadutils.factormap[pwmid], 'r')
calls = [line.strip().split()[:3] for line in handle]
handle.close()
macs = dict([(chrom, []) for chrom in utils.chromosomes[:22]])
[
macs[call[0]].append([int(call[1]), int(call[2])])
for call in calls if call[0] in utils.chromosomes[:22]
]
bsites = [
locs_tolearn[i] for i, p in enumerate(posterior[:, 1]) if p > 0.99
]
F, precision, sensitivity, ig = Fscore.Fscore(bsites, macs)
chipauc, tpr, positive, negative = compute_chip_auc(
chipreads, controlreads, logodds, macs, locs_tolearn)
print pwmid, model, sample, R, R2, chipauc, tpr, F, precision, sensitivity
output = {'footprint': footprints, \
'negbin': negbins, \
'prior': priors, \
'posterior': posteriors, \
'locations': locs_tolearn}
if sample is None:
handle = open(
"%s/cache/combined/pbmcentipede_%s_%s.pkl" %
(projpath, model, pwmid), 'w')
else:
handle = open(
"%s/cache/separate/pbmcentipede_%s_%s_%s.pkl" %
(projpath, model, pwmid, sample), 'w')
cPickle.Pickler(handle, protocol=2).dump(output)
handle.close()
readobj.close()
sequence.close()
def plotbound(pwmid, sample=None, cutk=0, pwmbase='transfac'):
import random
from matplotlib.backends.backend_pdf import PdfPages
bounds = [(1, 5), (5, 9), (9, 13), (13, np.inf)]
labels = ['1 - 5', '5 - 9', '9 - 13', '>13']
if pwmbase == 'transfac':
pwms = loadutils.transfac_pwms()
elif pwmbase == 'selex':
pwms = loadutils.selex_pwms()
dnaseobj = loadutils.Dnase(sample=sample)
chipseqobj = loadutils.ChipSeq('Gm12878', loadutils.factormap[pwmid])
mnaseobj = loadutils.Mnase(sample=sample)
indiv_idx = loadutils.read_individuals()
if sample in [None, 'Gm12878']:
sequence = loadutils.Sequence(sample, sample_idx=indiv_idx['NA18516'])
else:
sequence = loadutils.Sequence(sample, sample_idx=indiv_idx[sample])
key = [k for k, pwm in pwms.iteritems() if pwm['AC'] == pwmid][0]
bound_scores = []
bound_chipreads = []
unbound_chipreads = []
dnasemean_bound = []
mnasemean_bound = []
chiptotalreads = []
logodds = []
score = []
for bound in bounds:
# plot mean profile of all bound sites, stratified by PWM score
all_handle = loadutils.ZipFile("%s/cache/%s_locations_Q%.1f.txt.gz" %
(projpath, pwmid, 95.0))
if sample is None:
bound_handle = loadutils.ZipFile(
"%s/cache/combined/%s_%d_bound_Q%.1f.bed.gz" %
(projpath, pwmid, cutk, dhs))
else:
bound_handle = loadutils.ZipFile(
"%s/cache/separate/%s_%d_%s_bound_Q%.1f.bed.gz" %
(projpath, pwmid, cutk, sample, dhs))
all_locations = all_handle.read(threshold=bound)
blocs = bound_handle.read(threshold=bound)
bound_locations = [
loc[:5] for loc in blocs if float(loc[5]) >= np.log10(99)
]
if len(all_locations) > 2 * len(bound_locations):
all_locations = random.sample(all_locations,
2 * len(bound_locations))
unbound_locations = list(
set(all_locations).difference(set(bound_locations)))
chiptotalreads.extend([int(loc[-1]) for loc in blocs])
logodds.extend([float(loc[-2]) for loc in blocs])
score.extend([float(loc[-3]) for loc in blocs])
# load DNase and MNase reads
print bound, len(bound_locations), len(unbound_locations)
x, y = aggregate(bound_locations, dnaseobj)
dnasemean_bound.append(x)
mnasemean_bound.append(aggregate(bound_locations, mnaseobj))
# Total ChipSeq read counts
chipreads = chipseqobj.getreads(bound_locations)
bound_chipreads.extend(chipreads)
chipreads = chipseqobj.getreads(unbound_locations)
unbound_chipreads.extend(chipreads)
chiptotalreads = np.array(chiptotalreads)
logodds = np.array(logodds)
score = np.array(score)
if sample is None:
title = pwms[key]['NA']
tag = "_%s_%d_Q%.1f.pdf" % (pwmid, cutk, dhs)
dnaseprofilefile = "%s/fig/dnaseprofile%s" % (projpath, tag)
mnaseprofilefile = "%s/fig/mnaseprofile%s" % (projpath, tag)
chipdistfile = "%s/fig/chipdist%s" % (projpath, tag)
scatterfile = "%s/fig/scatter%s" % (projpath, tag)
scoreposfile = "%s/fig/scoreposition%s" % (projpath, tag)
posagreefile = "%s/fig/posagreement%s.pdf" % (projpath, tag)
else:
title = "%s / %s" % (pwms[key]['NA'], sample)
tag = "_short_%s_%d_%s_Q%.1f" % (pwmid, cutk, sample, dhs)
dnaseprofilefile = "%s/fig/dnaseprofile%s.pdf" % (projpath, tag)
mnaseprofilefile = "%s/fig/mnaseprofile%s.pdf" % (projpath, tag)
chipdistfile = "%s/fig/chipdist%s.pdf" % (projpath, tag)
scatterfile = "%s/fig/scatter%s.pdf" % (projpath, tag)
scoreposfile = "%s/fig/scoreposition%s.pdf" % (projpath, tag)
posagreefile = "%s/fig/posagreement%s.pdf" % (projpath, tag)
figure = viz.plot_dnaseprofile(dnasemean_bound,
labels,
motiflen=len(pwms[key]['motif']),
title=title)
figure.savefig(dnaseprofilefile, dpi=300, format='pdf')
figure = viz.plot_mnaseprofile(mnasemean_bound,
labels,
motiflen=len(pwms[key]['motif']),
title=title)
figure.savefig(mnaseprofilefile, dpi=300, format='pdf')
figure = viz.plot_chipseq_distribution(bound_chipreads,
unbound_chipreads,
title=title)
figure.savefig(chipdistfile, bbox_inches=0, dpi=300, format='pdf')
figure = viz.plot_chipseq_posterior_correlation(chiptotalreads,
logodds,
score,
title=title)
figure.savefig(scatterfile, bbox_inches=0, dpi=300, format='pdf')
dnaseobj.close()
mnaseobj.close()
chipseqobj.close()
sequence.close()
def decode(pwmid, sample, cutk=0, pwmbase='transfac', pos_threshold=np.log10(99), chipseq=False):
import centipede
import millipede
import centipede_pbm as pbmcentipede
if sample in [None,'Gm12878','Gm12878All']:
sequence = loadutils.Sequence(sample)
else:
indiv_idx = loadutils.read_individuals()
if pwmbase=='transfac':
pwms = loadutils.transfac_pwms()
elif pwmbase=='selex':
pwms = loadutils.selex_pwms()
motif = [val['motif'] for val in pwms.itervalues() if val['AC']==pwmid][0]
sequence = loadutils.Sequence(sample, sample_idx=indiv_idx[sample])
if cutk!=0:
sequence.set_cutrate(sample=sample, k=cutk)
# use output from Centipede run
# 0 = Py code, 1 = R code
if sample is None:
handle = open("%s/cache/combined/pbmcentipede_short_%s.pkl"%(projpath,pwmid),'r')
else:
handle = open("%s/cache/separate/pbmcentipede_short_%s_%s.pkl"%(projpath,pwmid,sample),'r')
output = cPickle.load(handle)
handle.close()
if cutk==0:
idx = 0
elif cutk==2:
idx = 1
elif cutk==4:
idx = 2
footprint = output['footprint'][idx]
negbinparams = output['negbin'][idx]
prior = output['prior'][idx][0]
dhsprior = output['prior'][idx][1]
if sample in ['Gm12878','Gm12878All']:
location_file = "%s/cache/%s_locationsGm12878_Q%.1f.txt.gz"%(projpath,pwmid,dhs)
else:
location_file = "%s/cache/%s_locations_Q%.1f.txt.gz"%(projpath,pwmid,dhs)
# check file size
pipe = subprocess.Popen("zcat %s | wc -l"%location_file, stdout=subprocess.PIPE, shell=True)
Ns = int(pipe.communicate()[0].strip())
try:
chipobj = loadutils.ChipSeq('Gm12878',loadutils.factormap[pwmid])
except:
pass
readobj = loadutils.Dnase(sample=sample)
readhandle = loadutils.ZipFile(location_file)
loops = Ns/batch
if sample is None:
handle = gzip.open("%s/cache/combined/%s_short_bound.bed.gz"%(projpath,pwmid),'wb')
else:
handle = gzip.open("%s/cache/separate/%s_%d_%s_short_bound.bed.gz"%(projpath,pwmid,cutk,sample),'wb')
towrite = ['Chr','Start','Stop','Strand','PwmScore','LogPosOdds','LogPriorOdds','MultLikeRatio','NegBinLikeRatio','ChipseqReads']
handle.write('\t'.join(towrite)+'\n')
totalreads = []
for n in xrange(loops):
starttime = time.time()
# read locations from file
locations = readhandle.read(chunk=batch)
if sample not in [None,'Gm12878','Gm12878All']:
# compute scores at locations for specific sample
locations = sequence.get_scores(locations, motif)
locations = sequence.filter_mappability(locations, width=max([200,L/2]))
# read in Dnase read data for locations
dnasereads, locations, subscores = readobj.getreads(locations, width=max([200,L/2]))
subscores = np.array(subscores).astype('float')
subscores = subscores.reshape(subscores.size,1)
dnasetotal = dnasereads.sum(1)
print len(locations)
if chipseq:
chipreads = chipobj.getreads(locations, width=max([200,L/2]))
else:
chipreads = None
# set null footprint distribution
if cutk==0:
null = np.ones((1,L),dtype=float)/L
else:
null = sequence.getnull(locations, width=L/2)
if L<400:
dnasereads = np.hstack((dnasereads[:,100-L/4:100+L/4],dnasereads[:,300-L/4:300+L/4]))
# if cutk==0:
logodds = centipede.decode(dnasereads, dnasetotal, null, subscores, footprint, negbinparams[0], negbinparams[1], prior, dhsprior, chipreads=chipreads, damp=damp)
# elif cutk==2:
# posterior = pbmcentipede.decode(reads, chipreads, subscores, footprint[1:], negbinparams[0], negbinparams[1], prior)
if not chipseq:
try:
chipreads = chipobj.get_total_reads(locations, width=400)
ignore = [loc.extend(['%.3f'%pos[0],'%.3f'%pos[1],'%.3f'%pos[2],'%.3f'%pos[3],'%d'%c]) \
for loc,pos,c in zip(locations,logodds,chipreads)]
except NameError:
ignore = [loc.extend(['%.3f'%pos[0],'%.3f'%pos[1],'%.3f'%pos[2],'%.3f'%pos[3]]) \
for loc,pos in zip(locations,logodds)]
locations = [loc for loc in locations if len(loc)>5]
ignore = [handle.write('\t'.join(elem)+'\n') for elem in locations]
print time.time()-starttime
remain = Ns-loops*batch
locations = readhandle.read(chunk=remain)
if sample not in [None,'Gm12878','Gm12878All']:
# compute scores at locations for specific sample
locations = sequence.get_scores(locations, motif)
locations = sequence.filter_mappability(locations, width=max([200,L/2]))
dnasereads, locations, subscores = readobj.getreads(locations, width=max([200,L/2]))
subscores = np.array(subscores)
subscores = subscores.reshape(subscores.size,1)
dnasetotal = dnasereads.sum(1)
if chipseq:
chipreads = chipobj.get_total_reads(locations, width=200)
else:
chipreads = None
# set null footprint distribution
if cutk==0:
null = np.ones((1,L),dtype=float)/L
else:
null = sequence.getnull(locations, width=L/2)
if L<400:
dnasereads = np.hstack((dnasereads[:,100-L/4:100+L/4],dnasereads[:,300-L/4:300+L/4]))
logodds = centipede.decode(dnasereads, dnasetotal, null, subscores, footprint, negbinparams[0], negbinparams[1], prior, dhsprior, chipreads=chipreads, damp=damp)
if not chipseq:
try:
chipreads = chipobj.get_total_reads(locations, width=400)
ignore = [loc.extend(['%.3f'%pos[0],'%.3f'%pos[1],'%.3f'%pos[2],'%.3f'%pos[3],'%d'%c]) \
for loc,pos,c in zip(locations,logodds,chipreads)]
except NameError:
ignore = [loc.extend(['%.3f'%pos[0],'%.3f'%pos[1],'%.3f'%pos[2],'%.3f'%pos[3]]) \
for loc,pos in zip(locations,logodds)]
locations = [loc for loc in locations if len(loc)>5]
ignore = [handle.write('\t'.join(elem)+'\n') for elem in locations]
readobj.close()
chipobj.close()
readhandle.close()
handle.close()
sequence.close()
def infer(pwmid, sample, pwm_thresh=8, pwmbase='transfac', chipseq=False):
import centipede_pbm as centipede
model = 'modelC'
if pwmbase=='transfac':
pwms = loadutils.transfac_pwms()
elif pwmbase=='selex':
pwms = loadutils.selex_pwms()
motif = [val['motif'] for val in pwms.itervalues() if val['AC']==pwmid][0]
if sample in [None,'Gm12878','Gm12878All']:
sequence = loadutils.Sequence(sample)
else:
indiv_idx = loadutils.read_individuals()
sequence = loadutils.Sequence(sample, sample_idx=indiv_idx[sample])
if sample in ['Gm12878','Gm12878All']:
location_file = "/mnt/lustre/home/anilraj/histmod/cache/%s_locationsGm12878_Q%.1f.txt.gz"%(pwmid,dhs)
else:
location_file = "/mnt/lustre/home/anilraj/histmod/cache/%s_locations_Q%.1f.txt.gz"%(pwmid,dhs)
# check file size
pipe = subprocess.Popen("zcat %s | wc -l"%location_file, stdout=subprocess.PIPE, shell=True)
Ns = int(pipe.communicate()[0].strip())
# load scores
alllocations = []
pwm_cutoff = pwm_thresh+1
while len(alllocations)<100:
pwm_cutoff = pwm_cutoff - 1
handle = loadutils.ZipFile(location_file)
alllocations = handle.read(threshold=pwm_cutoff)
handle.close()
print "PWM Cutoff = %d"%pwm_cutoff
# subsample locations, if too many
if len(alllocations)>100000:
scores = np.array([loc[-1] for loc in alllocations]).astype(float)
indices = np.argsort(scores)[-100000:]
alllocations = [alllocations[index] for index in indices]
print "Num of sites for learning, with pwm threshold of %d for %s = %d"%(pwm_thresh, pwmid, len(alllocations))
if sample in [None,'Gm12878','Gm12878All']:
locs_tolearn = alllocations
else:
# compute scores for specific sample at these locations
starttime = time.time()
locs_tolearn = sequence.get_scores(alllocations, motif)
print len(locs_tolearn), time.time()-starttime
# filter mappability
print "filtering out unmappable sites ..."
locs_tolearn = sequence.filter_mappability(locs_tolearn, width=max([200,L/2]))
# load reads and locations
print "loading dnase reads ..."
readobj = loadutils.Dnase(sample=sample)
dnasereads, locs_tolearn, subscores = readobj.getreads(locs_tolearn, remove_outliers=True, width=max([200,L/2]))
subscores = np.array(subscores)
subscores = subscores.reshape(subscores.size,1)
dnasetotal = dnasereads.sum(1)
print "Num of mappable sites for learning for %s = %d"%(pwmid,len(locs_tolearn))
if chipseq:
chipobj = loadutils.ChipSeq('Gm12878',loadutils.factormap[pwmid])
chipreads = chipobj.get_total_reads(locs_tolearn, width=200)
chipobj.close()
else:
chipreads = None
if L<400:
dnasereads = np.hstack((dnasereads[:,100-L/4:100+L/4],dnasereads[:,300-L/4:300+L/4]))
locs_tolearn = [list(loc) for loc in locs_tolearn]
footprints = []
priors = []
negbins = []
posteriors = []
null = np.ones((1,L),dtype=float)*1./L
posterior, footprint, negbinparams, prior = centipede.EM(dnasereads, dnasetotal, subscores, null, model=model, restarts=2)
posteriors.append(posterior)
footprints.append(footprint)
negbins.append(negbinparams)
priors.append(prior)
chipobj = loadutils.ChipSeq('Gm12878',loadutils.factormap[pwmid])
controlobj = loadutils.ChipSeq('Gm12878',loadutils.controlmap[pwmid])
chipreads = chipobj.get_total_reads(locs_tolearn, width=400)
controlreads = controlobj.get_total_reads(locs_tolearn, width=200)
chipobj.close()
controlobj.close()
for posterior in posteriors:
logodds = np.log(posterior[:,1]/posterior[:,0])
logodds[logodds==np.inf] = logodds[logodds!=np.inf].max()
logodds[logodds==-np.inf] = logodds[logodds!=-np.inf].min()
R = stats.pearsonr(logodds, np.sqrt(chipreads))
R2 = stats.pearsonr(np.sqrt(dnasetotal), np.sqrt(chipreads))
handle = open('/mnt/lustre/home/anilraj/histmod/cache/chipseq_peaks/%s_peaks.bed'%loadutils.factormap[pwmid],'r')
calls = [line.strip().split()[:3] for line in handle]
handle.close()
macs = dict([(chrom,[]) for chrom in utils.chromosomes[:22]])
[macs[call[0]].append([int(call[1]),int(call[2])]) for call in calls if call[0] in utils.chromosomes[:22]]
bsites = [locs_tolearn[i] for i,p in enumerate(posterior[:,1]) if p>0.99]
F, precision, sensitivity, ig = Fscore.Fscore(bsites, macs)
chipauc, tpr, positive, negative = compute_chip_auc(chipreads, controlreads, logodds, macs, locs_tolearn)
print pwmid, model, sample, R, R2, chipauc, tpr, F, precision, sensitivity
output = {'footprint': footprints, \
'negbin': negbins, \
'prior': priors, \
'posterior': posteriors, \
'locations': locs_tolearn}
if sample is None:
handle = open("%s/cache/combined/pbmcentipede_%s_%s.pkl"%(projpath,model,pwmid),'w')
else:
handle = open("%s/cache/separate/pbmcentipede_%s_%s_%s.pkl"%(projpath,model,pwmid,sample),'w')
cPickle.Pickler(handle, protocol=2).dump(output)
handle.close()
readobj.close()
sequence.close()
def plotbound(pwmid, sample=None, cutk=0, pwmbase='transfac'):
import random
from matplotlib.backends.backend_pdf import PdfPages
bounds = [(1,5),(5,9),(9,13),(13,np.inf)]
labels = ['1 - 5', '5 - 9', '9 - 13', '>13']
if pwmbase=='transfac':
pwms = loadutils.transfac_pwms()
elif pwmbase=='selex':
pwms = loadutils.selex_pwms()
dnaseobj = loadutils.Dnase(sample=sample)
chipseqobj = loadutils.ChipSeq('Gm12878',loadutils.factormap[pwmid])
mnaseobj = loadutils.Mnase(sample=sample)
indiv_idx = loadutils.read_individuals()
if sample in [None,'Gm12878']:
sequence = loadutils.Sequence(sample, sample_idx=indiv_idx['NA18516'])
else:
sequence = loadutils.Sequence(sample, sample_idx=indiv_idx[sample])
key = [k for k,pwm in pwms.iteritems() if pwm['AC']==pwmid][0]
bound_scores = []
bound_chipreads = []
unbound_chipreads = []
dnasemean_bound = []
mnasemean_bound = []
chiptotalreads = []
logodds = []
score = []
for bound in bounds:
# plot mean profile of all bound sites, stratified by PWM score
all_handle = loadutils.ZipFile("%s/cache/%s_locations_Q%.1f.txt.gz"%(projpath,pwmid,95.0))
if sample is None:
bound_handle = loadutils.ZipFile("%s/cache/combined/%s_%d_bound_Q%.1f.bed.gz"%(projpath,pwmid,cutk,dhs))
else:
bound_handle = loadutils.ZipFile("%s/cache/separate/%s_%d_%s_bound_Q%.1f.bed.gz"%(projpath,pwmid,cutk,sample,dhs))
all_locations = all_handle.read(threshold=bound)
blocs = bound_handle.read(threshold=bound)
bound_locations = [loc[:5] for loc in blocs if float(loc[5])>=np.log10(99)]
if len(all_locations)>2*len(bound_locations):
all_locations = random.sample(all_locations, 2*len(bound_locations))
unbound_locations = list(set(all_locations).difference(set(bound_locations)))
chiptotalreads.extend([int(loc[-1]) for loc in blocs])
logodds.extend([float(loc[-2]) for loc in blocs])
score.extend([float(loc[-3]) for loc in blocs])
# load DNase and MNase reads
print bound, len(bound_locations), len(unbound_locations)
x,y = aggregate(bound_locations, dnaseobj)
dnasemean_bound.append(x)
mnasemean_bound.append(aggregate(bound_locations, mnaseobj))
# Total ChipSeq read counts
chipreads = chipseqobj.getreads(bound_locations)
bound_chipreads.extend(chipreads)
chipreads = chipseqobj.getreads(unbound_locations)
unbound_chipreads.extend(chipreads)
chiptotalreads = np.array(chiptotalreads)
logodds = np.array(logodds)
score = np.array(score)
if sample is None:
title = pwms[key]['NA']
tag = "_%s_%d_Q%.1f.pdf"%(pwmid,cutk,dhs)
dnaseprofilefile = "%s/fig/dnaseprofile%s"%(projpath,tag)
mnaseprofilefile = "%s/fig/mnaseprofile%s"%(projpath,tag)
chipdistfile = "%s/fig/chipdist%s"%(projpath,tag)
scatterfile = "%s/fig/scatter%s"%(projpath,tag)
scoreposfile = "%s/fig/scoreposition%s"%(projpath,tag)
posagreefile = "%s/fig/posagreement%s.pdf"%(projpath,tag)
else:
title = "%s / %s"%(pwms[key]['NA'], sample)
tag = "_short_%s_%d_%s_Q%.1f"%(pwmid,cutk,sample,dhs)
dnaseprofilefile = "%s/fig/dnaseprofile%s.pdf"%(projpath,tag)
mnaseprofilefile = "%s/fig/mnaseprofile%s.pdf"%(projpath,tag)
chipdistfile = "%s/fig/chipdist%s.pdf"%(projpath,tag)
scatterfile = "%s/fig/scatter%s.pdf"%(projpath,tag)
scoreposfile = "%s/fig/scoreposition%s.pdf"%(projpath,tag)
posagreefile = "%s/fig/posagreement%s.pdf"%(projpath,tag)
figure = viz.plot_dnaseprofile(dnasemean_bound, labels, motiflen=len(pwms[key]['motif']), title=title)
figure.savefig(dnaseprofilefile, dpi=300, format='pdf')
figure = viz.plot_mnaseprofile(mnasemean_bound, labels, motiflen=len(pwms[key]['motif']), title=title)
figure.savefig(mnaseprofilefile, dpi=300, format='pdf')
figure = viz.plot_chipseq_distribution(bound_chipreads, unbound_chipreads, title=title)
figure.savefig(chipdistfile, bbox_inches=0, dpi=300, format='pdf')
figure = viz.plot_chipseq_posterior_correlation(chiptotalreads, logodds, score, title=title)
figure.savefig(scatterfile, bbox_inches=0, dpi=300, format='pdf')
dnaseobj.close()
mnaseobj.close()
chipseqobj.close()
sequence.close()
