def is_empty(vcf):
"""
Return True if the vcf file contains at least one variant, this doesn't use pysam
:param vcf: Path to possibly-gzipped vcf file
"""
for line in smartfile(vcf):
if line and not line.startswith('#'):
return False
return True
def batch_variants(vcf, max_batch_size=1000, min_safe_dist=2000):
"""
Given a list of variants, group them into batches such that no batch contains two variants
whose start positions are within min_safe_dist bases of each other
:param vcf: Filename of VCF file containing variants to batch
:param max_batch_size: Maximum number of variants per batch
:param min_safe_dist: Min permissible distance between two variants in batch
:return: List of VCF files containing subsets of variants
"""
batches = []
header = []
for x in smartfile(vcf):
if x.startswith('#'):
header.append(x)
else:
break
name = strip_extensions(os.path.basename(vcf), ['gz','vcf'])
vars = list(pysam.VariantFile(vcf))
while vars:
var = vars.pop(0)
unfilled_batches = [b for b in batches if len(b) < max_batch_size]
found = False
for b in unfilled_batches:
if canadd(var, b, max_batch_size, min_safe_dist=min_safe_dist):
b.append(var)
found = True
break
if not found:
# Need to make a new batch for this variant, it doesn't fit anywhere
batches.append([var])
# return batches
files = []
for i, batch in enumerate(batches, 1):
batchname = '{}.batch{:03d}.'.format(name, i) + randstr() + ".vcf"
with open(batchname, 'w') as out:
for x in header:
out.write(x)
for x in batch:
out.write(str(x))
files.append(batchname)
return files
def sort_vcf(vcf, conf):
tmpfile = vcf.replace(".vcf", ".sort" + randstr() + ".vcf").replace(".gz", "")
vars = []
ofh = open(tmpfile, "w")
for line in smartfile(vcf):
if line.startswith('#'):
ofh.write(line)
else:
vars.append(line.split('\t'))
for var in sorted(vars, cmp=var_comp):
ofh.write('\t'.join(var))
ofh.close()
return bgz_tabix(tmpfile, conf)
def set_genotypes(orig_vcf, newGT, region, conf, compress_result=True):
"""
Create a new VCF file that is identical to the given VCF, except that all GT info fields are set to 'newGT'
"""
# FIXME: Update to use pysam.VariantFile
newvcf = '{}.gtmod.{}.vcf'.format(strip_extensions(orig_vcf, ['gz','vcf']), randstr())
ofh = open(newvcf, 'w')
for line in smartfile(orig_vcf):
if line.startswith('#'):
ofh.write(line)
else:
toks = line.split('\t')
chrom = toks[0]
start = int(toks[1])
# FIXME: Interval intersection logic off? stop=start+rlen is not taken into consideration.
if region is not None and (chrom != region.chrom or start < region.start or start >= region.stop):
ofh.write(line)
continue
if len(toks) < 10:
ofh.write(line)
else:
if ',' in toks[4]:
raise GTModException('Cant set GT for multi-alt variants.')
if toks[4] == VCF_MISSING:
infoitems = ['0']
else:
infoitems = [newGT]
if ':' in toks[9]:
infoitems.extend(toks[9].strip().split(':')[1:])
newinfo = ':'.join(infoitems)
ofh.write('\t'.join(toks[0:9] + [newinfo]) + '\n')
ofh.close()
if compress_result:
newvcf = bgz_tabix(newvcf, conf)
return newvcf