def __call__(self, **kw):
assembly_id = kw.get('assembly') or None
assembly = genrep.Assembly(assembly_id)
tinput = track(kw.get('track'), chrmeta=assembly.chrmeta)
try:
thPromot = int(kw.get("promoter"))
except (ValueError, TypeError):
thPromot = prom_def
try:
thInter = int(kw.get("intergenic"))
except (ValueError, TypeError):
thInter = inter_def
try:
thUTR = int(kw.get("UTR"))
except (ValueError, TypeError):
thUTR = utr_def
output = self.temporary_path(fname=tinput.name+'_annotated.txt')
_fields = tinput.fields+['gene', 'location_type', 'distance']
tout = track(output, format='txt', fields=_fields)
tout.make_header("#"+"\t".join(tout.fields))
for chrom in assembly.chrnames:
tout.write(getNearestFeature(
tinput.read(selection=chrom),
assembly.gene_track(chrom),
thPromot, thInter, thUTR), mode='append')
tout.close()
self.new_file(output, 'table')
return self.display_time()
def test_getNearestFeature(self):
features = fstream([('chrII', 14795327, 14798367)],
fields=['chr', 'start', 'end'])
expected = [
('chrII', 14795327, 14798367, 'Y54E2A.12|tbc-20_Y54E2A.11|eif-3.B',
'Promot_Included', '28_0')
]
annotations = self.assembly.gene_track(chromlist=['chrII'])
res = list(getNearestFeature(features, annotations))
self.assertItemsEqual(res, expected)
def all_snps(ex, chrom, vcfs, bams, outall, assembly, sample_names, mincov, minsnp,
logfile=sys.stdout, debugfile=sys.stderr):
"""For a given chromosome, returns a summary file containing all SNPs identified
in at least one of the samples.
Each row contains: chromosome id, SNP position, reference base, SNP base (with proportions)
:param chrom: (str) chromosome name.
:param outall: (str) name of the file that will contain the list of all SNPs.
:param sample_names: (list of str) list of sample names.
:param mincov: (int) Minimum number of reads supporting an SNP at a position for it to be considered. [5]
:param minsnp: (int) Minimum percentage of reads supporting the SNP for it to be returned.
N.B.: Effectively, half of it on each strand for diploids. [40]
"""
ploidy = assembly.ploidy
allsnps = []
nsamples = len(sample_names)
sorder = range(len(sample_names))
current = [None]*nsamples
bam_tracks = [track(v,format='bam') for k,v in sorted(bams.items())]
vcf_handles = [open(v) for k,v in sorted(vcfs.items())]
for i,vh in enumerate(vcf_handles):
line = '#'
while line and line[0]=='#':
line = vh.readline()
current[i] = parse_vcf(line)
lastpos = 0
pos = -1
while any(current):
current_pos = [int(x[0][1]) if x else sys.maxint for x in current]
pos = min(current_pos)
if pos == sys.maxint: break
current_snp_idx = set(i for i in range(nsamples) if current_pos[i]==pos)
current_snps = ["0"]*nsamples
for i in current_snp_idx:
general,snp_info,sample_stats = current[i]
chrbam = general[0]
ref = general[2]
current_snps[i] = filter_snp(general,snp_info,sample_stats, mincov,minsnp,ploidy)
# If there were still snp called at this position after filtering
if any(current_snps[i] not in ("-","0") for i in current_snp_idx):
for i in set(range(nsamples))-current_snp_idx:
for coverage in bam_tracks[sorder[i]].coverage((chrom,pos-1,pos)):
if coverage[-1] > 0:
current_snps[i] = '-' # '/'.join([ref]*ploidy)
if pos != lastpos: # indel can be located at the same position as an SNP
allsnps.append((chrom,pos-1,pos,ref)+tuple(current_snps))
lastpos = pos
for i in current_snp_idx:
current[i] = parse_vcf(vcf_handles[i].readline())
for f in vcf_handles: f.close()
for b in bam_tracks: b.close()
logfile.write(" Annotate all SNPs\n"); logfile.flush()
snp_read = FeatureStream(allsnps, fields=['chr','start','end','name']+sample_names)
try:
annotation = assembly.gene_track(chrom)
annotated_stream = gm_stream.getNearestFeature(snp_read,annotation,
thresholdPromot=3000, thresholdInter=3000, thresholdUTR=10)
except:
annotated_stream = snp_read
logfile.write(" Write all SNPs\n"); logfile.flush()
with open(outall,"a") as fout:
for snp in annotated_stream:
# snp: ('chrV',154529, 154530,'T','A (50% of 10)','A (80% of 10)',
# 'YER002W|NOP16_YER001W|MNN1','Upstream_Included','2271_1011')
fout.write('\t'.join(str(x) for x in (snp[0],)+snp[2:])+'\n') # remove start coord (0-based)
return allsnps
def test_getNearestFeature(self):
features = fstream([('chrII',14795327,14798367)], fields=['chr','start','end'])
expected = [('chrII',14795327, 14798367, 'Y54E2A.12|tbc-20_Y54E2A.11|eif-3.B', 'Promot_Included', '28_0')]
annotations = self.assembly.gene_track(chromlist=['chrII'])
res = list(getNearestFeature(features,annotations))
self.assertItemsEqual(res,expected)
def all_snps(ex, chrom, vcfs, bams, outall, assembly, headerfile, sample_names, mincov, minsnp,
logfile=sys.stdout, debugfile=sys.stderr, via='local'):
"""For a given chromosome, returns a summary file containing all SNPs identified
in at least one of the samples.
Each row contains: chromosome id, SNP position, reference base, SNP base (with proportions)
:param chrom: (str) chromosome name.
:param vcfs: (dict) vcf files for each sample, dictionary keys are group ids.
:param bams: (dict) bamfiles organized like the vcf files.
:param outall: (str) name of the file that will contain the list of all SNPs.
:param assembly: (genrep.Assembly) assembly for the fasta files and ploidy value.
:param headerfile: (string) name of file with substitute bam header to match the fasta files.
:param sample_names: (list of str) list of sample names.
:param mincov: (int) minimum number of reads supporting an SNP at a position for it to be considered. [5]
:param minsnp: (int) minimum percentage of reads supporting the SNP for it to be returned.
N.B.: Effectively, half of it on each strand for diploids. [40]
"""
ploidy = assembly.ploidy
reffasta = assembly.fasta_by_chrom[chrom]
allsnps = []
nsamples = len(sample_names)
sorder = range(len(sample_names))
current = [None]*nsamples
#####
if nsamples > 1:
poslist = set()
bamchrom = None
for vf in vcfs.values():
with open(vf) as vh:
for line in vh:
if (not line) or line[0]=='#': continue
line = line.strip().split('\t')
if bamchrom is None: bamchrom = line[0]
poslist.add( int(line[1]) )
snplist = unique_filename_in()
with open(snplist,"w") as snpfh:
snpfh.write("\n".join("%s\t%i" %(bamchrom,pos) for pos in sorted(poslist)))
pilejobs = []
vcfs2 = {}
for gid, bamfile in bams.iteritems():
vcfs2[gid] = unique_filename_in()
pilejobs.append( pileup.nonblocking(ex, bams[gid], reffasta,
step="list", bedfile=snplist, header=headerfile,
via=via, stdout=vcfs2[gid]) )
[job.wait() for job in pilejobs]
else:
vcfs2 = vcfs
#####
bam_tracks = [track(v,format='bam') for k,v in sorted(bams.items())]
vcf_handles = [open(v) for k,v in sorted(vcfs2.items())]
for i,vh in enumerate(vcf_handles):
line = '#'
while line and line[0]=='#':
line = vh.readline()
current[i] = parse_vcf(line)
lastpos = 0
pos = -1
while any(current):
current_pos = [int(x[0][1]) if x else sys.maxint for x in current]
pos = min(current_pos)
if pos == sys.maxint: break
current_snp_idx = set(i for i in range(nsamples) if current_pos[i]==pos)
current_snps = ["0"]*nsamples
for i in current_snp_idx:
general,snp_info,sample_stats = current[i]
chrbam = general[0]
ref = general[2]
current_snps[i] = filter_snp(general,snp_info,sample_stats, mincov,minsnp,ploidy)
# If there were still snp called at this position after filtering
if any(current_snps[i] not in ("-","0") for i in current_snp_idx):
for i in set(range(nsamples))-current_snp_idx:
for coverage in bam_tracks[sorder[i]].coverage((chrom,pos-1,pos)):
if coverage[-1] > 0:
current_snps[i] = '-' # '/'.join([ref]*ploidy)
if pos != lastpos: # indel can be located at the same position as an SNP
allsnps.append((chrom,pos-1,pos,ref)+tuple(current_snps))
lastpos = pos
for i in current_snp_idx:
current[i] = parse_vcf(vcf_handles[i].readline())
for f in vcf_handles: f.close()
for b in bam_tracks: b.close()
logfile.write(" Annotate all SNPs\n"); logfile.flush()
snp_read = FeatureStream(allsnps, fields=['chr','start','end','name']+sample_names)
try:
annotation = assembly.gene_track(chrom)
annotated_stream = gm_stream.getNearestFeature(snp_read,annotation,
thresholdPromot=3000, thresholdInter=3000, thresholdUTR=10)
except:
annotated_stream = snp_read
logfile.write(" Write all SNPs\n"); logfile.flush()
with open(outall,"a") as fout:
for snp in annotated_stream:
# snp: ('chrV',154529, 154530,'T','A (50% of 10)','A (80% of 10)',
# 'YER002W|NOP16_YER001W|MNN1','Upstream_Included','2271_1011')
fout.write('\t'.join(str(x) for x in (snp[0],)+snp[2:])+'\n') # remove start coord (0-based)
return allsnps
def chipseq_workflow( ex, job_or_dict, assembly, script_path='', logfile=sys.stdout, via='lsf' ):
"""Runs a chipseq workflow over bam files obtained by mapseq. Will optionally run ``macs`` and 'run_deconv'.
:param ex: a 'bein' execution environment to run jobs in,
:param job_or_dict: a 'Frontend' 'job' object, or a dictionary with key 'groups', 'files' and 'options' if applicable,
:param assembly: a genrep.Assembly object,
:param script_path: only needed if 'run_deconv' is in the job options, must point to the location of the R scripts.
Defaults ``macs`` parameters (overriden by ``job_or_dict['options']['macs_args']``) are set as follows:
* ``'-bw'``: 200 ('bandwith')
* ``'-m'``: 10,100 ('minimum and maximum enrichments relative to background or control')
The enrichment bounds will be computed from a Poisson threshold *T*, if available, as *(min(30,5*(T+1)),50*(T+1))*.
Returns a tuple of a dictionary with keys *group_id* from the job groups, *macs* and *deconv* if applicable and values file description dictionaries and a dictionary of *group_ids* to *names* used in file descriptions.
"""
options = {}
if logfile is None: logfile = sys.stdout
if isinstance(job_or_dict,frontend.Job):
options = job_or_dict.options
groups = job_or_dict.groups
mapseq_files = job_or_dict.files
elif isinstance(job_or_dict,dict) and 'groups' in job_or_dict:
if 'options' in job_or_dict:
options = job_or_dict['options']
groups = job_or_dict['groups']
for gid in groups.keys():
if not('name' in groups[gid]):
groups[gid]['name'] = gid
mapseq_files = job_or_dict.get('files',{})
else:
raise TypeError("job_or_dict must be a frontend. Job object or a dictionary with key 'groups'.")
merge_strands = int(options.get('merge_strands',-1))
suffixes = ["fwd","rev"]
peak_deconvolution = options.get('peak_deconvolution',False)
if isinstance(peak_deconvolution,basestring):
peak_deconvolution = peak_deconvolution.lower() in ['1','true','t']
run_meme = options.get('run_meme',False)
if isinstance(run_meme,basestring):
run_meme = run_meme.lower() in ['1','true','t']
macs_args = options.get('macs_args',["--bw","200"])
b2w_args = options.get('b2w_args',[])
if not(isinstance(mapseq_files,dict)):
raise TypeError("Mapseq_files must be a dictionary.")
tests = []
controls = []
names = {'tests': [], 'controls': []}
read_length = []
p_thresh = {}
for gid,mapped in mapseq_files.iteritems():
group_name = groups[gid]['name']
if not(isinstance(mapped,dict)):
raise TypeError("Mapseq_files values must be dictionaries with keys *run_ids* or 'bam'.")
if 'bam' in mapped:
mapped = {'_': mapped}
futures = {}
ptruns = []
for k in mapped.keys():
if not 'libname' in mapped[k]:
mapped[k]['libname'] = group_name+"_"+str(k)
if not 'stats' in mapped[k]:
futures[k] = mapseq.bamstats.nonblocking( ex, mapped[k]["bam"], via=via )
if mapped[k].get('poisson_threshold',-1)>0:
ptruns.append(mapped[k]['poisson_threshold'])
if len(ptruns)>0:
p_thresh['group_name'] = sum(ptruns)/len(ptruns)
for k in futures.keys():
mapped[k]['stats'] = f.wait()
if len(mapped)>1:
bamfile = mapseq.merge_bam(ex, [m['bam'] for m in mapped.values()])
else:
bamfile = mapped.values()[0]['bam']
if groups[gid]['control']:
controls.append(bamfile)
names['controls'].append((gid,group_name))
else:
tests.append(bamfile)
names['tests'].append((gid,group_name))
read_length.append(mapped.values()[0]['stats']['read_length'])
genome_size = mapped.values()[0]['stats']['genome_size']
if len(controls)<1:
controls = [None]
names['controls'] = [(0,None)]
logfile.write("Starting MACS.\n");logfile.flush()
processed = {'macs': add_macs_results( ex, read_length, genome_size,
tests, ctrlbam=controls, name=names,
poisson_threshold=p_thresh,
macs_args=macs_args, via=via ) }
logfile.write("Done MACS.\n");logfile.flush()
peak_list = {}
chrlist = assembly.chrmeta
## select only peaks with p-val <= 1e-0.6 = .25 => score = -10log10(p) >= 6
_select = {'score':(6,sys.maxint)}
_fields = ['chr','start','end','name','score']
for i,name in enumerate(names['tests']):
if len(names['controls']) < 2:
ctrl = (name,names['controls'][0])
macsbed = track(processed['macs'][ctrl]+"_summits.bed",
chrmeta=chrlist, fields=_fields).read(selection=_select)
else:
macsbed = concatenate([apply(track(processed['macs'][(name,x)]+"_summits.bed",
chrmeta=chrlist, fields=_fields).read(selection=_select),
'name', lambda __n,_n=xn: "%s:%i" %(__n,_n))
for xn,x in enumerate(names['controls'])])
##############################
macs_neighb = neighborhood( macsbed, before_start=150, after_end=150 )
peak_list[name] = unique_filename_in()+".sql"
macs_final = track( peak_list[name], chrmeta=chrlist,
info={'datatype':'qualitative'},
fields=['start','end','name','score'] )
macs_final.write(fusion(macs_neighb),clip=True)
macs_final.close()
##############################
merged_wig = {}
options['read_extension'] = int(options.get('read_extension') or read_length[0])
if options['read_extension'] < 1: options['read_extension'] = read_length[0]
make_wigs = merge_strands >= 0 or options['read_extension']>100
if options['read_extension'] > 100: options['read_extension'] = 50
for gid,mapped in mapseq_files.iteritems():
# if groups[gid]['control']: continue
group_name = groups[gid]['name']
wig = []
for m in mapped.values():
if make_wigs or not('wig' in m) or len(m['wig'])<2:
output = mapseq.parallel_density_sql( ex, m["bam"], assembly.chrmeta,
nreads=m["stats"]["total"],
merge=-1, read_extension=options['read_extension'],
convert=False,
b2w_args=b2w_args, via=via )
wig.append(dict((s,output+s+'.sql') for s in suffixes))
else:
wig.append(m['wig'])
if len(wig) > 1:
merged_wig[group_name] = dict((s,merge_sql(ex, [x[s] for x in wig], via=via))
for s in suffixes)
else:
merged_wig[group_name] = wig[0]
if peak_deconvolution:
##############################
def _filter_deconv( stream, pval ):
ferr = re.compile(r';FERR=([\d\.]+)$')
return FeatureStream( ((x[0],)+((x[2]+x[1])/2-150,(x[2]+x[1])/2+150)+x[3:]
for x in stream
if "FERR=" in x[3] and float(ferr.search(x[3]).groups()[0]) <= pval),
fields=stream.fields )
##############################
processed['deconv'] = {}
for name in names['tests']:
logfile.write(name[1]+" deconvolution.\n");logfile.flush()
if len(names['controls']) < 2:
ctrl = (name,names['controls'][0])
macsbed = processed['macs'][ctrl]+"_peaks.bed"
else:
macsbed = intersect_many_bed( ex, [processed['macs'][(name,x)]+"_peaks.bed"
for x in names['controls']], via=via )
deconv = run_deconv( ex, merged_wig[name[1]], macsbed, assembly.chrmeta,
options['read_extension'], script_path, via=via )
peak_list[name] = unique_filename_in()+".bed"
trbed = track(deconv['peaks']).read()
with track(peak_list[name], chrmeta=chrlist, fields=trbed.fields) as bedfile:
bedfile.write(fusion(_filter_deconv(trbed,0.65)))
ex.add(deconv['peaks'],
description=set_file_descr(name[1]+'_peaks.sql', type='sql',
step='deconvolution', groupId=name[0]))
ex.add(deconv['profile'],
description=set_file_descr(name[1]+'_deconv.sql', type='sql',
step='deconvolution', groupId=name[0]))
bigwig = unique_filename_in()
try:
convert(deconv['profile'],(bigwig,"bigWig"))
ex.add(bigwig,
description=set_file_descr(name[1]+'_deconv.bw', type='bigWig',
ucsc='1', step='deconvolution',
groupId=name[0]))
except OSError as e:
logfile.write(str(e));logfile.flush()
ex.add(deconv['pdf'],
description=set_file_descr(name[1]+'_deconv.pdf', type='pdf',
step='deconvolution', groupId=name[0]))
processed['deconv'][name] = deconv
##############################
def _join_macs( stream, xlsl, _f ):
def _macs_row(_s):
for _p in _s:
for _n in _p[3].split("|"):
if len(xlsl) == 1:
nb = int(_n.split(";")[0][13:]) if _n[:3] == "ID=" else int(_n[10:])
yield _p+xlsl[0][nb-1][1:]
else:
nb = _n.split(";")[0][13:] if _n[:3] == "ID=" else _n[10:]
nb = nb.split(":")
yield _p+xlsl[int(nb[1])][int(nb[0])-1][1:]
return FeatureStream( _macs_row(stream), fields=_f )
##############################
peakfile_list = []
for name, plist in peak_list.iteritems():
ptrack = track(plist,chrmeta=chrlist,fields=["chr","start","end","name","score"])
peakfile = unique_filename_in()
xlsh, xlsl = parse_MACS_xls([processed['macs'][(name,_c)]+"_peaks.xls" for _c in names['controls']])
try:
###### if assembly doesn't have annotations, we skip the "getNearestFeature" but still go through "_join_macs"
assembly.gene_track()
_fields = ['chr','start','end','name','score','gene','location_type','distance']\
+["MACS_%s"%h for h in xlsh[1:5]]+xlsh[5:]
peakout = track(peakfile, format='txt', chrmeta=chrlist, fields=_fields)
peakout.make_header("#"+"\t".join(['chromosome','start','end','info','peak_height','gene(s)','location_type','distance']+_fields[8:]))
for chrom in assembly.chrnames:
_feat = assembly.gene_track(chrom)
peakout.write(_join_macs(getNearestFeature(ptrack.read(selection=chrom),_feat),
xlsl, _fields), mode='append')
except ValueError:
_fields = ['chr','start','end','name','score']+["MACS_%s"%h for h in xlsh[1:5]]+xlsh[5:]
peakout = track(peakfile, format='txt', chrmeta=chrlist, fields=_fields)
peakout.make_header("#"+"\t".join(['chromosome','start','end','info','peak_height']+_fields[8:]))
for chrom in assembly.chrnames:
peakout.write(_join_macs(ptrack.read(selection=chrom), xlsl, _fields), mode='append')
peakout.close()
gzipfile(ex,peakfile)
peakfile_list.append(track(peakfile+".gz", format='txt', fields=_fields))
ex.add(peakfile+".gz",
description=set_file_descr(name[1]+'_annotated_peaks.txt.gz',type='text',
step='annotation',groupId=name[0]))
stracks = [track(wig,info={'name':name+"_"+st})
for name,wigdict in merged_wig.iteritems() for st,wig in wigdict.iteritems()]
tablefile = unique_filename_in()
with open(tablefile,"w") as _tf:
_pnames = ["MACS_%s_vs_%s" %(_s[1],_c[1]) if _c[1] else "MACS_%s" %_s[1]
for _s in names['tests'] for _c in names['controls']]
_tf.write("\t".join(['#chromosome','start','end',]+_pnames+[s.name for s in stracks])+"\n")
#### need to do something about peak origin (split names, write to separate columns?)
for chrom in assembly.chrnames:
pk_lst = [apply(pt.read(chrom,fields=['chr','start','end','name']),
'name', lambda __n,_n=npt: "%s:%i" %(__n,_n))
for npt,pt in enumerate(peakfile_list)]
features = fusion(concatenate(pk_lst, fields=['chr','start','end','name'],
remove_duplicates=True, group_by=['chr','start','end']))
sread = [sig.read(chrom) for sig in stracks]
quantifs = score_by_feature(sread, features, method='sum')
nidx = quantifs.fields.index('name')
_ns = len(tests)
_nc = len(controls)
with open(tablefile,"a") as _tf:
for row in quantifs:
pcols = ['']*_ns*_nc
_rnsplit = row[nidx].split(":")
_n1 = _rnsplit[0]
_k = 0
while ( _k < len(_rnsplit)-1-int(_nc>1) ):
if _nc > 1:
_k += 2
_n2 = _rnsplit[_k-1]
_n = _rnsplit[_k].split("|")
pcols[int(_n[0])*_nc+int(_n2)] = _n1
else:
_k += 1
_n = _rnsplit[_k].split("|")
pcols[int(_n[0])] = _n1
_n1 = "|".join(_n[1:])
_tf.write("\t".join(str(tt) for tt in row[:nidx]+tuple(pcols)+row[nidx+1:])+"\n")
gzipfile(ex,tablefile)
ex.add(tablefile+".gz",
description=set_file_descr('Combined_peak_quantifications.txt.gz',type='text',
step='summary'))
if run_meme:
from bbcflib.motif import parallel_meme
logfile.write("Starting MEME.\n");logfile.flush()
processed['meme'] = parallel_meme( ex, assembly,
peak_list.values(), name=peak_list.keys(),
chip=True, meme_args=['-meme-nmotifs','4','-meme-mod','zoops'],
via=via )
return processed
def chipseq_workflow(ex,
job_or_dict,
assembly,
script_path='',
logfile=sys.stdout,
via='lsf'):
"""Runs a chipseq workflow over bam files obtained by mapseq. Will optionally run ``macs`` and 'run_deconv'.
:param ex: a 'bein' execution environment to run jobs in,
:param job_or_dict: a 'Frontend' 'job' object, or a dictionary with key 'groups', 'files' and 'options' if applicable,
:param assembly: a genrep.Assembly object,
:param script_path: only needed if 'run_deconv' is in the job options, must point to the location of the R scripts.
Defaults ``macs`` parameters (overriden by ``job_or_dict['options']['macs_args']``) are set as follows:
* ``'-bw'``: 200 ('bandwith')
* ``'-m'``: 10,100 ('minimum and maximum enrichments relative to background or control')
The enrichment bounds will be computed from a Poisson threshold *T*, if available, as *(min(30,5*(T+1)),50*(T+1))*.
Returns a tuple of a dictionary with keys *group_id* from the job groups, *macs* and *deconv* if applicable and values file description dictionaries and a dictionary of *group_ids* to *names* used in file descriptions.
"""
options = {}
if logfile is None: logfile = sys.stdout
if isinstance(job_or_dict, frontend.Job):
options = job_or_dict.options
groups = job_or_dict.groups
mapseq_files = job_or_dict.files
elif isinstance(job_or_dict, dict) and 'groups' in job_or_dict:
if 'options' in job_or_dict:
options = job_or_dict['options']
groups = job_or_dict['groups']
for gid in groups.keys():
if not ('name' in groups[gid]):
groups[gid]['name'] = gid
mapseq_files = job_or_dict.get('files', {})
else:
raise TypeError(
"job_or_dict must be a frontend. Job object or a dictionary with key 'groups'."
)
merge_strands = int(options.get('merge_strands', -1))
suffixes = ["fwd", "rev"]
peak_deconvolution = options.get('peak_deconvolution', False)
if isinstance(peak_deconvolution, basestring):
peak_deconvolution = peak_deconvolution.lower() in ['1', 'true', 't']
run_meme = options.get('run_meme', False)
if isinstance(run_meme, basestring):
run_meme = run_meme.lower() in ['1', 'true', 't']
macs_args = options.get('macs_args', ["--bw", "200"])
b2w_args = options.get('b2w_args', [])
if not (isinstance(mapseq_files, dict)):
raise TypeError("Mapseq_files must be a dictionary.")
tests = []
controls = []
names = {'tests': [], 'controls': []}
read_length = []
p_thresh = {}
for gid, mapped in mapseq_files.iteritems():
group_name = groups[gid]['name']
if not (isinstance(mapped, dict)):
raise TypeError(
"Mapseq_files values must be dictionaries with keys *run_ids* or 'bam'."
)
if 'bam' in mapped:
mapped = {'_': mapped}
futures = {}
ptruns = []
for k in mapped.keys():
if not 'libname' in mapped[k]:
mapped[k]['libname'] = group_name + "_" + str(k)
if not 'stats' in mapped[k]:
futures[k] = mapseq.bamstats.nonblocking(ex,
mapped[k]["bam"],
via=via)
if mapped[k].get('poisson_threshold', -1) > 0:
ptruns.append(mapped[k]['poisson_threshold'])
if len(ptruns) > 0:
p_thresh['group_name'] = sum(ptruns) / len(ptruns)
for k in futures.keys():
mapped[k]['stats'] = f.wait()
if len(mapped) > 1:
bamfile = mapseq.merge_bam(ex, [m['bam'] for m in mapped.values()])
else:
bamfile = mapped.values()[0]['bam']
if groups[gid]['control']:
controls.append(bamfile)
names['controls'].append((gid, group_name))
else:
tests.append(bamfile)
names['tests'].append((gid, group_name))
read_length.append(mapped.values()[0]['stats']['read_length'])
genome_size = mapped.values()[0]['stats']['genome_size']
if len(controls) < 1:
controls = [None]
names['controls'] = [(0, None)]
logfile.write("Starting MACS.\n")
logfile.flush()
processed = {
'macs':
add_macs_results(ex,
read_length,
genome_size,
tests,
ctrlbam=controls,
name=names,
poisson_threshold=p_thresh,
macs_args=macs_args,
via=via)
}
logfile.write("Done MACS.\n")
logfile.flush()
peak_list = {}
chrlist = assembly.chrmeta
## select only peaks with p-val <= 1e-0.6 = .25 => score = -10log10(p) >= 6
_select = {'score': (6, sys.maxint)}
_fields = ['chr', 'start', 'end', 'name', 'score']
for i, name in enumerate(names['tests']):
if len(names['controls']) < 2:
ctrl = (name, names['controls'][0])
macsbed = track(processed['macs'][ctrl] + "_summits.bed",
chrmeta=chrlist,
fields=_fields).read(selection=_select)
else:
macsbed = concatenate([
apply(track(processed['macs'][(name, x)] + "_summits.bed",
chrmeta=chrlist,
fields=_fields).read(selection=_select),
'name',
lambda __n, _n=xn: "%s:%i" % (__n, _n))
for xn, x in enumerate(names['controls'])
])
##############################
macs_neighb = neighborhood(macsbed, before_start=150, after_end=150)
peak_list[name] = unique_filename_in() + ".sql"
macs_final = track(peak_list[name],
chrmeta=chrlist,
info={'datatype': 'qualitative'},
fields=['start', 'end', 'name', 'score'])
macs_final.write(fusion(macs_neighb), clip=True)
macs_final.close()
##############################
merged_wig = {}
options['read_extension'] = int(
options.get('read_extension') or read_length[0])
if options['read_extension'] < 1:
options['read_extension'] = read_length[0]
make_wigs = merge_strands >= 0 or options['read_extension'] > 100
if options['read_extension'] > 100: options['read_extension'] = 50
for gid, mapped in mapseq_files.iteritems():
# if groups[gid]['control']: continue
group_name = groups[gid]['name']
wig = []
for m in mapped.values():
if make_wigs or not ('wig' in m) or len(m['wig']) < 2:
output = mapseq.parallel_density_sql(
ex,
m["bam"],
assembly.chrmeta,
nreads=m["stats"]["total"],
merge=-1,
read_extension=options['read_extension'],
convert=False,
b2w_args=b2w_args,
via=via)
wig.append(dict((s, output + s + '.sql') for s in suffixes))
else:
wig.append(m['wig'])
if len(wig) > 1:
merged_wig[group_name] = dict(
(s, merge_sql(ex, [x[s] for x in wig], via=via))
for s in suffixes)
else:
merged_wig[group_name] = wig[0]
if peak_deconvolution:
##############################
def _filter_deconv(stream, pval):
ferr = re.compile(r';FERR=([\d\.]+)$')
return FeatureStream(
((x[0], ) + ((x[2] + x[1]) / 2 - 150,
(x[2] + x[1]) / 2 + 150) + x[3:]
for x in stream if "FERR=" in x[3]
and float(ferr.search(x[3]).groups()[0]) <= pval),
fields=stream.fields)
##############################
processed['deconv'] = {}
for name in names['tests']:
logfile.write(name[1] + " deconvolution.\n")
logfile.flush()
if len(names['controls']) < 2:
ctrl = (name, names['controls'][0])
macsbed = processed['macs'][ctrl] + "_peaks.bed"
else:
macsbed = intersect_many_bed(ex, [
processed['macs'][(name, x)] + "_peaks.bed"
for x in names['controls']
],
via=via)
deconv = run_deconv(ex,
merged_wig[name[1]],
macsbed,
assembly.chrmeta,
options['read_extension'],
script_path,
via=via)
peak_list[name] = unique_filename_in() + ".bed"
trbed = track(deconv['peaks']).read()
with track(peak_list[name], chrmeta=chrlist,
fields=trbed.fields) as bedfile:
bedfile.write(fusion(_filter_deconv(trbed, 0.65)))
ex.add(deconv['peaks'],
description=set_file_descr(name[1] + '_peaks.sql',
type='sql',
step='deconvolution',
groupId=name[0]))
ex.add(deconv['profile'],
description=set_file_descr(name[1] + '_deconv.sql',
type='sql',
step='deconvolution',
groupId=name[0]))
bigwig = unique_filename_in()
try:
convert(deconv['profile'], (bigwig, "bigWig"))
ex.add(bigwig,
description=set_file_descr(name[1] + '_deconv.bw',
type='bigWig',
ucsc='1',
step='deconvolution',
groupId=name[0]))
except OSError as e:
logfile.write(str(e))
logfile.flush()
ex.add(deconv['pdf'],
description=set_file_descr(name[1] + '_deconv.pdf',
type='pdf',
step='deconvolution',
groupId=name[0]))
processed['deconv'][name] = deconv
##############################
def _join_macs(stream, xlsl, _f):
def _macs_row(_s):
for _p in _s:
for _n in _p[3].split("|"):
if len(xlsl) == 1:
nb = int(
_n.split(";")[0][13:]) if _n[:3] == "ID=" else int(
_n[10:])
yield _p + xlsl[0][nb - 1][1:]
else:
nb = _n.split(
";")[0][13:] if _n[:3] == "ID=" else _n[10:]
nb = nb.split(":")
yield _p + xlsl[int(nb[1])][int(nb[0]) - 1][1:]
return FeatureStream(_macs_row(stream), fields=_f)
##############################
peakfile_list = []
for name, plist in peak_list.iteritems():
ptrack = track(plist,
chrmeta=chrlist,
fields=["chr", "start", "end", "name", "score"])
peakfile = unique_filename_in()
xlsh, xlsl = parse_MACS_xls([
processed['macs'][(name, _c)] + "_peaks.xls"
for _c in names['controls']
])
try:
###### if assembly doesn't have annotations, we skip the "getNearestFeature" but still go through "_join_macs"
assembly.gene_track()
_fields = ['chr','start','end','name','score','gene','location_type','distance']\
+["MACS_%s"%h for h in xlsh[1:5]]+xlsh[5:]
peakout = track(peakfile,
format='txt',
chrmeta=chrlist,
fields=_fields)
peakout.make_header("#" + "\t".join([
'chromosome', 'start', 'end', 'info', 'peak_height', 'gene(s)',
'location_type', 'distance'
] + _fields[8:]))
for chrom in assembly.chrnames:
_feat = assembly.gene_track(chrom)
peakout.write(_join_macs(
getNearestFeature(ptrack.read(selection=chrom), _feat),
xlsl, _fields),
mode='append')
except ValueError:
_fields = ['chr', 'start', 'end', 'name', 'score'
] + ["MACS_%s" % h for h in xlsh[1:5]] + xlsh[5:]
peakout = track(peakfile,
format='txt',
chrmeta=chrlist,
fields=_fields)
peakout.make_header("#" + "\t".join(
['chromosome', 'start', 'end', 'info', 'peak_height'] +
_fields[8:]))
for chrom in assembly.chrnames:
peakout.write(_join_macs(ptrack.read(selection=chrom), xlsl,
_fields),
mode='append')
peakout.close()
gzipfile(ex, peakfile)
peakfile_list.append(
track(peakfile + ".gz", format='txt', fields=_fields))
ex.add(peakfile + ".gz",
description=set_file_descr(name[1] + '_annotated_peaks.txt.gz',
type='text',
step='annotation',
groupId=name[0]))
stracks = [
track(wig, info={'name': name + "_" + st})
for name, wigdict in merged_wig.iteritems()
for st, wig in wigdict.iteritems()
]
tablefile = unique_filename_in()
with open(tablefile, "w") as _tf:
_pnames = [
"MACS_%s_vs_%s" % (_s[1], _c[1]) if _c[1] else "MACS_%s" % _s[1]
for _s in names['tests'] for _c in names['controls']
]
_tf.write("\t".join([
'#chromosome',
'start',
'end',
] + _pnames + [s.name for s in stracks]) + "\n")
#### need to do something about peak origin (split names, write to separate columns?)
for chrom in assembly.chrnames:
pk_lst = [
apply(pt.read(chrom, fields=['chr', 'start', 'end', 'name']),
'name',
lambda __n, _n=npt: "%s:%i" % (__n, _n))
for npt, pt in enumerate(peakfile_list)
]
features = fusion(
concatenate(pk_lst,
fields=['chr', 'start', 'end', 'name'],
remove_duplicates=True,
group_by=['chr', 'start', 'end']))
sread = [sig.read(chrom) for sig in stracks]
quantifs = score_by_feature(sread, features, method='sum')
nidx = quantifs.fields.index('name')
_ns = len(tests)
_nc = len(controls)
with open(tablefile, "a") as _tf:
for row in quantifs:
pcols = [''] * _ns * _nc
_rnsplit = row[nidx].split(":")
_n1 = _rnsplit[0]
_k = 0
while (_k < len(_rnsplit) - 1 - int(_nc > 1)):
if _nc > 1:
_k += 2
_n2 = _rnsplit[_k - 1]
_n = _rnsplit[_k].split("|")
pcols[int(_n[0]) * _nc + int(_n2)] = _n1
else:
_k += 1
_n = _rnsplit[_k].split("|")
pcols[int(_n[0])] = _n1
_n1 = "|".join(_n[1:])
_tf.write("\t".join(
str(tt)
for tt in row[:nidx] + tuple(pcols) + row[nidx + 1:]) +
"\n")
gzipfile(ex, tablefile)
ex.add(tablefile + ".gz",
description=set_file_descr('Combined_peak_quantifications.txt.gz',
type='text',
step='summary'))
if run_meme:
from bbcflib.motif import parallel_meme
logfile.write("Starting MEME.\n")
logfile.flush()
processed['meme'] = parallel_meme(
ex,
assembly,
peak_list.values(),
name=peak_list.keys(),
chip=True,
meme_args=['-meme-nmotifs', '4', '-meme-mod', 'zoops'],
via=via)
return processed
def all_snps(ex,
chrom,
vcfs,
bams,
outall,
assembly,
headerfile,
sample_names,
mincov,
minsnp,
logfile=sys.stdout,
debugfile=sys.stderr,
via='local'):
"""For a given chromosome, returns a summary file containing all SNPs identified
in at least one of the samples.
Each row contains: chromosome id, SNP position, reference base, SNP base (with proportions)
:param chrom: (str) chromosome name.
:param vcfs: (dict) vcf files for each sample, dictionary keys are group ids.
:param bams: (dict) bamfiles organized like the vcf files.
:param outall: (str) name of the file that will contain the list of all SNPs.
:param assembly: (genrep.Assembly) assembly for the fasta files and ploidy value.
:param headerfile: (string) name of file with substitute bam header to match the fasta files.
:param sample_names: (list of str) list of sample names.
:param mincov: (int) minimum number of reads supporting an SNP at a position for it to be considered. [5]
:param minsnp: (int) minimum percentage of reads supporting the SNP for it to be returned.
N.B.: Effectively, half of it on each strand for diploids. [40]
"""
ploidy = assembly.ploidy
reffasta = assembly.fasta_by_chrom[chrom]
allsnps = []
nsamples = len(sample_names)
sorder = range(len(sample_names))
current = [None] * nsamples
#####
if nsamples > 1:
poslist = set()
bamchrom = None
for vf in vcfs.values():
with open(vf) as vh:
for line in vh:
if (not line) or line[0] == '#': continue
line = line.strip().split('\t')
if bamchrom is None: bamchrom = line[0]
poslist.add(int(line[1]))
snplist = unique_filename_in()
with open(snplist, "w") as snpfh:
snpfh.write("\n".join("%s\t%i" % (bamchrom, pos)
for pos in sorted(poslist)))
pilejobs = []
vcfs2 = {}
for gid, bamfile in bams.iteritems():
vcfs2[gid] = unique_filename_in()
pilejobs.append(
pileup.nonblocking(ex,
bams[gid],
reffasta,
step="list",
bedfile=snplist,
header=headerfile,
via=via,
stdout=vcfs2[gid]))
[job.wait() for job in pilejobs]
else:
vcfs2 = vcfs
#####
bam_tracks = [track(v, format='bam') for k, v in sorted(bams.items())]
vcf_handles = [open(v) for k, v in sorted(vcfs2.items())]
for i, vh in enumerate(vcf_handles):
line = '#'
while line and line[0] == '#':
line = vh.readline()
current[i] = parse_vcf(line)
lastpos = 0
pos = -1
while any(current):
current_pos = [int(x[0][1]) if x else sys.maxint for x in current]
pos = min(current_pos)
if pos == sys.maxint: break
current_snp_idx = set(i for i in range(nsamples)
if current_pos[i] == pos)
current_snps = ["0"] * nsamples
for i in current_snp_idx:
general, snp_info, sample_stats = current[i]
chrbam = general[0]
ref = general[2]
current_snps[i] = filter_snp(general, snp_info, sample_stats,
mincov, minsnp, ploidy)
# If there were still snp called at this position after filtering
if any(current_snps[i] not in ("-", "0") for i in current_snp_idx):
for i in set(range(nsamples)) - current_snp_idx:
for coverage in bam_tracks[sorder[i]].coverage(
(chrom, pos - 1, pos)):
if coverage[-1] > 0:
current_snps[i] = '-' # '/'.join([ref]*ploidy)
if pos != lastpos: # indel can be located at the same position as an SNP
allsnps.append((chrom, pos - 1, pos, ref) +
tuple(current_snps))
lastpos = pos
for i in current_snp_idx:
current[i] = parse_vcf(vcf_handles[i].readline())
for f in vcf_handles:
f.close()
for b in bam_tracks:
b.close()
logfile.write(" Annotate all SNPs\n")
logfile.flush()
snp_read = FeatureStream(allsnps,
fields=['chr', 'start', 'end', 'name'] +
sample_names)
try:
annotation = assembly.gene_track(chrom)
annotated_stream = gm_stream.getNearestFeature(snp_read,
annotation,
thresholdPromot=3000,
thresholdInter=3000,
thresholdUTR=10)
except:
annotated_stream = snp_read
logfile.write(" Write all SNPs\n")
logfile.flush()
with open(outall, "a") as fout:
for snp in annotated_stream:
# snp: ('chrV',154529, 154530,'T','A (50% of 10)','A (80% of 10)',
# 'YER002W|NOP16_YER001W|MNN1','Upstream_Included','2271_1011')
fout.write('\t'.join(str(x) for x in (snp[0], ) + snp[2:]) +
'\n') # remove start coord (0-based)
return allsnps
