def fastq_tile(fname, outbase, length, offset, gz=False, quiet=False):
fastq = FASTQ(fname)
outs = []
fnames = []
for read in fastq.fetch(quiet=quiet):
out_idx = 0
pos = 0
while pos + length < len(read.seq):
if len(outs) <= out_idx:
fobj, tmp, fn = _open_file(outbase, out_idx, gz, quiet)
outs.append(fobj)
fnames.append((tmp, fn))
read.subseq(pos, pos + length, comment="#tile:%s,%s" % (pos, pos + length)).write(outs[out_idx])
pos += offset
out_idx += 1
for out in outs:
out.close()
fastq.close()
for tmp, fname in fnames:
os.rename(tmp, fname)
def fastq_unmerge(combined_fname, out_template, gz=False):
outs = []
if gz:
outs.append(gzip.open('%s.1.fastq.gz' % out_template, 'w'))
else:
outs.append(open('%s.1.fastq' % out_template, 'w'))
outidx = 1
last_read = None
fq = FASTQ(combined_fname)
for read in fq.fetch():
if last_read and last_read.name == read.name:
outidx += 1
if len(outs) < outidx:
if gz:
outs.append(gzip.open('%s.%s.fastq.gz' % (out_template, outidx), 'w'))
else:
outs.append(open('%s.%s.fastq' % (out_template, outidx), 'w'))
read.write(outs[outidx - 1])
else:
outidx = 1
read.write(outs[0])
last_read = read
fq.close()
for out in outs:
out.close()
def fastq_split(fname, outbase, chunks, ignore_pairs=False, gz=False, count_fname=None, quiet=False):
fastq = FASTQ(fname)
if ignore_pairs:
is_paired = False
else:
is_paired = fastq.is_paired
outs = []
fnames = []
for i in xrange(chunks):
if gz:
fn = '%s.%s.fastq.gz' % (outbase, i + 1)
tmp = os.path.join(os.path.dirname(fn), '.tmp.%s' % os.path.basename(fn))
fnames.append((tmp, fn))
if not quiet:
sys.stderr.write('Output file: %s\n' % fn)
outs.append(gzip.open(tmp, 'w'))
else:
fn = '%s.%s.fastq' % (outbase, i + 1)
tmp = os.path.join(os.path.dirname(fn), '.tmp.%s' % os.path.basename(fn))
fnames.append((tmp, fn))
if not quiet:
sys.stderr.write('Output file: %s\n' % fn)
outs.append(open(tmp, 'w'))
i = chunks
last_name = None
for read in fastq.fetch(quiet=quiet):
if not is_paired:
i += 1
elif read.name != last_name:
i += 1
if i >= len(outs):
i = 0
last_name = read.name
read.write(outs[i])
for out in outs:
out.close()
fastq.close()
for tmp, fname in fnames:
os.rename(tmp, fname)
def testFQRead(self):
fq = StringIO.StringIO('''\
@foo
ACGTacgtACGT
+
CDEFGHIJKLMN
''')
out = StringIO.StringIO('')
ngsutils.fastq.convertqual.fastq_convertqual(FASTQ(fileobj=fq), out=out, quiet=True)
out.seek(0)
fqout = FASTQ(fileobj=out)
read = fqout.fetch().next()
self.assertEqual(read.name, 'foo')
self.assertEqual(read.seq, 'ACGTacgtACGT')
self.assertEqual(read.qual, "$%&'()*+,-./")
def assert_fastq_contains(self, base, args):
for tag in args:
valid = args[tag][0].split()
seq_qual = {}
if args[tag][1]:
for n, s, q in zip(valid, args[tag][1].split(), args[tag][2].split()):
seq_qual[n] = (s, q)
fq = FASTQ(base % tag)
count = 0
for read in fq.fetch():
if read.name in valid:
count += 1
if seq_qual:
self.assertEqual(seq_qual[read.name], (read.seq, read.qual))
else:
self.assertEqual('extra read in %s' % tag, read.name)
self.assertEqual(count, len(valid))
def testSplitUnpaired(self):
fname = os.path.join(os.path.dirname(__file__), 'test.fastq')
templ = os.path.join(os.path.dirname(__file__), 'test_templ')
ngsutils.fastq.split.fastq_split(fname, templ, 2, ignore_pairs=True, quiet=True)
self.assertTrue(os.path.exists('%s.1.fastq' % templ))
self.assertTrue(os.path.exists('%s.2.fastq' % templ))
fq1 = FASTQ('%s.1.fastq' % templ)
fq2 = FASTQ('%s.2.fastq' % templ)
names1 = [x.name for x in fq1.fetch(quiet=True)]
self.assertEqual(names1, ['foo', 'bar', 'baz'])
names2 = [x.name for x in fq2.fetch(quiet=True)]
self.assertEqual(names2, ['foo', 'bar', 'baz'])
fq1.close()
fq2.close()
os.unlink('%s.1.fastq' % templ)
os.unlink('%s.2.fastq' % templ)
outname = None
read1_fname = None
read2_fname = None
force = False
for arg in sys.argv[1:]:
if arg == "-f":
force = True
elif not outname:
if not force and os.path.exists(arg):
usage("Output file exists! (Use -f to force overwriting): %s" % arg)
outname = arg
elif not read1_fname and os.path.exists(arg):
read1_fname = arg
elif not read2_fname and os.path.exists(arg):
read2_fname = arg
if not outname or not read1_fname:
usage()
read1 = FASTQ(read1_fname)
read2 = FASTQ(read2_fname) if read2_fname else None
bam = pysam.Samfile(outname, "wb")
export_bam(bam, read1, read2)
bam.close()
read1.close()
read2.close()
if not os.path.exists(arg):
usage("File %s doesn't exist!" % arg)
fqname2 = arg
elif not outname1:
if os.path.exists(arg):
usage("File %s exists!" % arg)
outname1 = arg
elif not outname2:
if os.path.exists(arg):
usage("File %s exists!" % arg)
outname2 = arg
if not fqname1 or not fqname2 or not outname1 or not outname2:
usage()
fq1 = FASTQ(fqname1)
fq2 = FASTQ(fqname2)
if gz:
out1 = gzip.open(outname1, 'w')
out2 = gzip.open(outname2, 'w')
else:
out1 = open(outname1, 'w')
out2 = open(outname2, 'w')
total1, total2, matched = find_fastq_pairs(fq1, fq2, out1, out2)
print "Totals: %s, %s" % (total1, total2)
print "Proper pairs: %s" % matched
fq1.close()
fname = arg
if not fname or not filters_config:
usage()
discard = None
_d_file = None
if discard_fname:
_d_file = open(discard_fname, 'w')
def _callback(name):
_d_file.write('%s\n' % name[1:])
discard = _callback
fq = FASTQ(fname)
chain = FASTQReader(fq, veryverbose)
for config in filters_config:
if verbose:
sys.stderr.write(config[0].__name__)
sys.stderr.write('\t%s\n' % '\t'.join([str(x) for x in config[1:]]))
clazz = config[0]
opts = config[1:]
if clazz == QualFilter:
chain = clazz(chain, *opts, verbose=veryverbose, discard=discard, illumina=illumina)
else:
chain = clazz(chain, *opts, verbose=veryverbose, discard=discard)
def testSplitThree(self):
fname = os.path.join(os.path.dirname(__file__), 'test.fastq')
templ = os.path.join(os.path.dirname(__file__), 'test_templ')
ngsutils.fastq.split.fastq_split(fname, templ, 3, ignore_pairs=True, quiet=True)
self.assertTrue(os.path.exists('%s.1.fastq' % templ))
self.assertTrue(os.path.exists('%s.2.fastq' % templ))
self.assertTrue(os.path.exists('%s.3.fastq' % templ))
fq1 = FASTQ('%s.1.fastq' % templ)
fq2 = FASTQ('%s.2.fastq' % templ)
fq3 = FASTQ('%s.3.fastq' % templ)
names1 = [x.fullname for x in fq1.fetch(quiet=True)]
self.assertEqual(names1, ['foo /1', 'bar /2'])
names2 = [x.fullname for x in fq2.fetch(quiet=True)]
self.assertEqual(names2, ['foo /2', 'baz /1'])
names3 = [x.fullname for x in fq3.fetch(quiet=True)]
self.assertEqual(names3, ['bar /1', 'baz /2'])
fq1.close()
fq2.close()
fq3.close()
os.unlink('%s.1.fastq' % templ)
os.unlink('%s.2.fastq' % templ)
os.unlink('%s.3.fastq' % templ)
def find_fastq_pairs(fq1, fq2, out1, out2, tmpdir=None, quiet=False):
tmp1 = tempfile.NamedTemporaryFile(delete=False, prefix='.tmp', suffix='.gz', dir=tmpdir if tmpdir else os.path.dirname(fq1.fname))
tmp1_fname = tmp1.name
tmp1_out = gzip.GzipFile(fileobj=tmp1)
ngsutils.fastq.sort.fastq_sort(fq1, out=tmp1_out, tmpdir=tmpdir if tmpdir else os.path.dirname(fq1.fname))
tmp1_out.close()
tmp1.close()
tmp2 = tempfile.NamedTemporaryFile(delete=False, prefix='.tmp', suffix='.gz', dir=tmpdir if tmpdir else os.path.dirname(fq2.fname))
tmp2_fname = tmp2.name
tmp2_out = gzip.GzipFile(fileobj=tmp2)
ngsutils.fastq.sort.fastq_sort(fq2, out=tmp2_out, tmpdir=tmpdir if tmpdir else os.path.dirname(fq2.fname))
tmp2_out.close()
tmp2.close()
sys.stderr.write('Finding properly paired FASTQ reads...\n')
fq_tmp1 = FASTQ(tmp1_fname)
fq_tmp2 = FASTQ(tmp2_fname)
reader1 = fq_tmp1.fetch(quiet=quiet)
reader2 = fq_tmp2.fetch(quiet=True)
read1 = reader1.next()
read2 = reader2.next()
pairs = 0
discarded_1 = 0
discarded_2 = 0
while read1 and read2:
if read1.name == read2.name:
read1.write(out1)
read2.write(out2)
try:
read1 = reader1.next()
read2 = reader2.next()
except StopIteration:
break
pairs += 1
elif read1.name < read2.name:
discarded_1 += 1
try:
read1 = reader1.next()
except StopIteration:
break
else:
discarded_2 += 1
try:
read2 = reader2.next()
except StopIteration:
break
fq_tmp1.close()
fq_tmp2.close()
os.unlink(tmp1_fname)
os.unlink(tmp2_fname)
return pairs, discarded_1, discarded_2
read.clone(name=name, comment=comment).write(out)
def usage():
print __doc__
print """Usage: fastqutils merge {-slash} file1.fastq{.gz} file2.fastq{.gz} ...
-slash Split the read name at a '/' (Illumina paired format)
"""
sys.exit(1)
if __name__ == '__main__':
fnames = []
split_slashes = False
for arg in sys.argv[1:]:
if arg == '-slash':
split_slashes = True
elif os.path.exists(arg):
fnames.append(arg)
if len(fnames) < 2:
usage()
fastqs = [FASTQ(x) for x in fnames]
fastq_merge(fastqs, split_slashes)
for fq in fastqs:
fq.close()
def find_fastq_pairs(fq1, fq2, out1, out2, tmpdir=None, quiet=False):
tmp1 = tempfile.NamedTemporaryFile(
delete=False,
prefix='.tmp',
suffix='.gz',
dir=tmpdir if tmpdir else os.path.dirname(fq1.fname))
tmp1_fname = tmp1.name
tmp1_out = gzip.GzipFile(fileobj=tmp1)
ngsutils.fastq.sort.fastq_sort(
fq1,
out=tmp1_out,
tmpdir=tmpdir if tmpdir else os.path.dirname(fq1.fname))
tmp1_out.close()
tmp1.close()
tmp2 = tempfile.NamedTemporaryFile(
delete=False,
prefix='.tmp',
suffix='.gz',
dir=tmpdir if tmpdir else os.path.dirname(fq2.fname))
tmp2_fname = tmp2.name
tmp2_out = gzip.GzipFile(fileobj=tmp2)
ngsutils.fastq.sort.fastq_sort(
fq2,
out=tmp2_out,
tmpdir=tmpdir if tmpdir else os.path.dirname(fq2.fname))
tmp2_out.close()
tmp2.close()
sys.stderr.write('Finding properly paired FASTQ reads...\n')
fq_tmp1 = FASTQ(tmp1_fname)
fq_tmp2 = FASTQ(tmp2_fname)
reader1 = fq_tmp1.fetch(quiet=quiet)
reader2 = fq_tmp2.fetch(quiet=True)
read1 = reader1.next()
read2 = reader2.next()
pairs = 0
discarded_1 = 0
discarded_2 = 0
while read1 and read2:
if read1.name == read2.name:
read1.write(out1)
read2.write(out2)
try:
read1 = reader1.next()
read2 = reader2.next()
except StopIteration:
break
pairs += 1
elif read1.name < read2.name:
discarded_1 += 1
try:
read1 = reader1.next()
except StopIteration:
break
else:
discarded_2 += 1
try:
read2 = reader2.next()
except StopIteration:
break
fq_tmp1.close()
fq_tmp2.close()
os.unlink(tmp1_fname)
os.unlink(tmp2_fname)
return pairs, discarded_1, discarded_2
usage("File %s doesn't exist!" % arg)
fqname2 = arg
elif not outname1:
outname1 = arg
elif not outname2:
outname2 = arg
if not fqname1 or not fqname2 or not outname1 or not outname2:
usage()
if not force:
for fname in [outname1, outname2]:
if os.path.exists(fname):
usage("File %s exists!" % fname)
fq1 = FASTQ(fqname1)
fq2 = FASTQ(fqname2)
if gz:
out1 = gzip.open(outname1, 'w')
out2 = gzip.open(outname2, 'w')
else:
out1 = open(outname1, 'w')
out2 = open(outname2, 'w')
paired, discard_1, discard_2 = find_fastq_pairs(fq1, fq2, out1, out2,
tmpdir)
print "Proper pairs: %s" % paired
print "Discarded 1 : %s" % discard_1
print "Discarded 2 : %s" % discard_2
for read in fastq.fetch(quiet=quiet):
if include_comment:
out.write('%s%s%s\n' %
(read.name, ' ' if read.comment else '', read.comment))
else:
out.write('%s\n' % read.name)
def usage():
print __doc__
print "Usage: fastqutils names {-comment} filename.fastq{.gz}"
sys.exit(1)
if __name__ == '__main__':
fname = None
include_comment = False
for arg in sys.argv[1:]:
if arg == '-comment':
include_comment = True
elif os.path.exists(arg):
fname = arg
if not fname:
usage()
fq = FASTQ(fname)
export_names(fq, include_comment)
fq.close()