class ICell8FastqIterator(Iterator):
"""
Class for iterating over an iCell8 R1/R2 FASTQ-pair
The iterator returns a set of ICell8ReadPair
instances, for example:
>>> for pair in ICell8FastqIterator(fq1,fq2):
>>> print "-- R1: %s" % pair.r1
>>> print " R2: %s" % pair.r2
"""
def __init__(self,fqr1,fqr2):
"""
Create a new ICell8FastqIterator instance
Arguments:
fqr1 (str): path to the R1 FASTQ file
fqr2 (str): path to the R2 FASTQ
"""
self._read_count = 0
self._fqr1 = FastqIterator(fqr1)
self._fqr2 = FastqIterator(fqr2)
def next(self):
self._read_count += 1
r1 = self._fqr1.next()
r2 = self._fqr2.next()
try:
return ICell8ReadPair(r1,r2)
except Exception as ex:
print "Failed to create read pair:"
print "-- Read pair number: %d" % self._read_count
print "-- Read 1:\n%s" % r1
print "-- Read 2:\n%s" % r2
logging.critical("Failed to create read pair: %s" % ex)
raise ex
def from_fastq(self, fastq):
"""
Get statistics from a FASTQ file
Generates and stores statistics from a FASTQ file.
Arguments:
fastq (str): path to a FASTQ file (can be gzipped)
"""
# Initialise using first read for sequence length
quality_per_base = []
for read in FastqIterator(fastq):
for i in xrange(read.seqlen):
quality_per_base.append({})
for j in xrange(ord('!'), ord('I') + 1):
quality_per_base[i][chr(j)] = 0
break
# Iterate through fastq file and count quality scores
nreads = 0
for read in FastqIterator(fastq):
nreads += 1
for pos, q in enumerate(read.quality):
quality_per_base[pos][q] += 1
#print quality_per_base
# Median etc positions
# FIXME these are not correct if the list has an odd number of values!
median_pos = nreads / 2
q25_pos = median_pos / 2
q75_pos = median_pos + q25_pos
# For percentiles see http://stackoverflow.com/a/2753343/579925
# FIXME 10th/90th percentiles is a fudge here
p10_pos = nreads / 10
p90_pos = nreads - p10_pos
# For each base position determine stats
for pos, counts in enumerate(quality_per_base):
#print "Position: %d" % pos
# Expand to a list
scores = ''
for q in counts:
scores += q * counts[q]
# Sort into order
scores = ''.join(sorted(scores))
#print scores
# Get the mean (scores are Phred+33 encoded)
self.mean.append(
float(sum([(ord(q) - 33) for q in scores])) / nreads)
#print "Mean: %.2f" % self.mean[pos]
# Get the median etc
self.median.append(ord(scores[median_pos]) - 33)
self.q25.append(ord(scores[q25_pos]) - 33)
self.q75.append(ord(scores[q75_pos]) - 33)
self.p10.append(ord(scores[p10_pos]) - 33)
self.p90.append(ord(scores[p90_pos]) - 33)
def __init__(self,fqr1,fqr2):
"""
Create a new ICell8FastqIterator instance
Arguments:
fqr1 (str): path to the R1 FASTQ file
fqr2 (str): path to the R2 FASTQ
"""
self._read_count = 0
self._fqr1 = FastqIterator(fqr1)
self._fqr2 = FastqIterator(fqr2)
def collect_fastq_stats(fastq):
"""
Get barcode and distince UMI counts for Fastq file
Used by Icell8Stats to collect counts for each file
supplied.
Arguments:
fastq (str): path to Fastq file
Returns:
Tuple: tuple consisting of (fastq,counts,umis)
where 'fastq' is the path to the input Fastq
file, 'counts' is a dictionary with barcodes
as keys and read counts as values, and 'umis'
is a dictionary with barcodes as keys and
sets of UMIs as values.
"""
counts = {}
umis = {}
for r in FastqIterator(fastq):
r = ICell8Read1(r)
barcode = r.barcode
try:
counts[barcode] += 1
except KeyError:
counts[barcode] = 1
umi = r.umi
try:
umis[barcode].add(umi)
except KeyError:
umis[barcode] = set((umi, ))
return (fastq, counts, umis)
def pair_fastqs(fastqs):
"""
Automagically pair up FASTQ files
Given a list of FASTQ files, generate a list of R1/R2
pairs by examining the header for the first read in
each file.
Arguments:
fastqs (list): list of paths to FASTQ files which
will be paired.
Returns:
Tuple: pair of lists of the form (paired,unpaired),
where `paired` is a list of tuples consisting of
FASTQ R1/R2 pairs and `unpaired` is a list of
FASTQs which couldn't be paired.
"""
fq_pairs = []
seq_ids = {}
bad_files = []
for fq in [os.path.abspath(fq) for fq in fastqs]:
# Get header from first read
seq_id = None
with get_fastq_file_handle(fq) as fp:
for r in FastqIterator(fp=fp):
seq_id = r.seqid
break
if seq_id is None:
logging.debug("'Bad' file: %s" % fq)
bad_files.append(fq)
continue
fq_pair = None
for fq1 in seq_ids:
if seq_id.is_pair_of(seq_ids[fq1]):
# Found a pair
if seq_id.pair_id == '1':
fq_pair = (fq, fq1)
else:
fq_pair = (fq1, fq)
fq_pairs.append(fq_pair)
logging.debug("*** Paired: %s\n" " : %s" % fq_pair)
# Remove paired fastq
del (seq_ids[fq1])
break
if fq_pair is None:
# Unable to pair, store for now
logging.debug("Unpaired: %s" % fq)
seq_ids[fq] = seq_id
# Sort pairs into order
fq_pairs = sorted(fq_pairs, key=lambda x: x[0])
unpaired = sorted(list(seq_ids.keys()) + bad_files)
# Return paired and upaired fastqs
return (fq_pairs, unpaired)
def collect_fastq_stats(self,fastq):
"""
Get barcode and distinct UMI counts for Fastq file
This method can be called directly, but is
also invoked implicitly if its parent instance
is called.
Arguments:
fastq (str): path to Fastq file
Returns:
Tuple: tuple consisting of (fastq,counts,umis)
where 'fastq' is the path to the input Fastq
file, 'counts' is a dictionary with barcodes
as keys and read counts as values, and 'umis'
is a dictionary with barcodes as keys and
sets of UMIs as values.
"""
print("collect_fastq_stats: started: %s" % fastq)
try:
n = FastqReadCounter.zcat_wc(fastq)
print("%s: processing %d read%s" % (
os.path.basename(fastq),
n,('s' if n != 1 else '')))
counts = {}
umis = {}
progress = ProgressChecker(percent=5,total=n)
for i,r in enumerate(FastqIterator(fastq),start=1):
r = ICell8Read1(r)
barcode = r.barcode
try:
counts[barcode] += 1
except KeyError:
counts[barcode] = 1
umi = r.umi
try:
umis[barcode].add(umi)
except KeyError:
umis[barcode] = set((umi,))
if self._verbose:
if progress.check(i):
print("%s: %s: processed %d reads (%.1f%%)" %
(time.strftime("%Y%m%d.%H%M%S"),
os.path.basename(fastq),
i,progress.percent(i)))
except Exception as ex:
print("collect_fastq_stats: caught exception: '%s'" % ex)
raise Exception("collect_fastq_stats: %s: caught exception "
"'%s'",(fastq,ex))
print("collect_fastq_stats: returning: %s" % fastq)
return (fastq,counts,umis)
def get_read_number(fastq):
"""
Get the read number (1 or 2) from a Fastq file
Arguments:
fastq (str): path to a Fastq file
Returns:
Integer: read number (1 or 2) extracted from the first read.
"""
for r in FastqIterator(fastq):
seq_id = r.seqid
break
return int(seq_id.pair_id)
def count_barcodes(fastqs):
"""
Count the barcodes from multiple fastqs
"""
print "Reading in %s fastq%s" % (len(fastqs),
('' if len(fastqs) == 1 else 's'))
counts = BarcodeCounter()
for fq in fastqs:
print "%s" % os.path.basename(fq)
for r in FastqIterator(fq):
seq = r.seqid.index_sequence
lane = int(r.seqid.flowcell_lane)
counts.count_barcode(seq, lane)
return counts
def fastqiterator(fastq=None, fp=None):
"""
Return number of reads in a FASTQ file
Uses the FASTQFile.FastqIterator class to do the
counting.
Arguments:
fastq: fastq(.gz) file
fp: open file descriptor for fastq file
Returns:
Number of reads
"""
nreads = 0
for r in FastqIterator(fastq_file=fastq, fp=fp):
nreads += 1
return nreads
def assign_barcodes_single_end(fastq_in, fastq_out, n=5):
"""
Extract inline barcodes and assign to Fastq read headers
Strips the first n bases from each read of the input
FASTQ file and assigns it to the index sequence for that
read in the output file.
If the supplied output file name ends with '.gz' then it
will be gzipped.
Arguments:
fastq_in (str): input FASTQ file (can be gzipped)
fastq_out (str): output FASTQ file (will be gzipped if
ending with '.gz')
n (integer): number of bases to extract and assign as
index sequence (default: 5)
Returns:
Integer: number of reads processed.
"""
if fastq_out.endswith('.gz'):
fp = gzip.GzipFile(filename=fastq_out, mode='wb')
else:
fp = open(fastq_out, 'w')
print("Processing reads from %s" % fastq_in)
nread = 0
for read in FastqIterator(fastq_in):
# Extract new barcode sequence
barcode = read.sequence[:n]
# Truncate sequence and quality accordingly
sequence = read.sequence[n:]
quality = read.quality[n:]
# Assign new values and write to output
read.seqid.index_sequence = barcode
read.sequence = sequence
read.quality = quality
fp.write("%s\n" % read)
nread += 1
fp.close()
print("Finished (%d reads processed)" % nread)
return nread
def reads_per_lane(fastq=None, fp=None):
"""
Return counts of reads in each lane of FASTQ file
Uses the FASTQFile.FastqIterator class to do the
counting, with counts split by lane.
Arguments:
fastq: fastq(.gz) file
fp: open file descriptor for fastq file
Returns:
Dictionary where keys are lane numbers (as integers)
and values are number of reads in that lane.
"""
nreads = {}
for r in FastqIterator(fastq_file=fastq, fp=fp):
lane = int(r.seqid.flowcell_lane)
try:
nreads[lane] += 1
except KeyError:
nreads[lane] = 1
return nreads
ndatabases_mammalian = len(nohits_mammalian)
ndatabases_contaminants = len(nohits_contaminants)
print "'nohits' tags: '%s' and '%s'" % (nohits_mammalian,
nohits_contaminants)
# Output filtered FASTQ pair
fqr1_out = os.path.basename(strip_ext(fqr1,'.fastq')) \
+ '.filtered.fastq'
fqr2_out = os.path.basename(strip_ext(fqr2,'.fastq')) \
+ '.filtered.fastq'
output_fqs = OutputFiles(base_dir=out_dir)
output_fqs.open('fqr1', fqr1_out)
output_fqs.open('fqr2', fqr2_out)
# Filter the iCell8 read pairs against the tagged reads
for pair, pref, contam in izip(ICell8FastqIterator(fqr1, fqr2),
FastqIterator(mammalian_tagged_fq),
FastqIterator(contaminants_tagged_fq)):
# Get the tags
pref_tag = extract_fastq_screen_tag(pref)
contam_tag = extract_fastq_screen_tag(contam)
# Check number of databases are consistent
if len(pref_tag) != ndatabases_mammalian:
logging.critical("Mismatch in mammalian tag: "
"len('%s') != len('%s')" %
(pref_tag, nohits_mammalian))
sys.exit(1)
if len(contam_tag) != ndatabases_contaminants:
logging.critical("Mismatch in contaminant tag: "
"len('%s') != len('%s')" %
(contam_tag, nohits_contaminants))
sys.exit(1)
def get_sequence_lengths(fastq, outfile=None, show_progress=False, limit=None):
"""
Get sequence lengths and masking statistics for Fastq
Returns a dictionary with the following keys:
- fastq: the Fastq file that metrics were calculated
from
- nreads: total number of reads processed
- nreads_masked: number of reads that are completely
masked (i.e. consist only of 'N's)
- nreads_padded: number of partially masked reads
(i.e. contain trailing 'N's)
- frac_reads_masked: fraction of the processed reads
which are masked
- frac_reads_padded: fraction of the processed reads
which are padded
- min_length: minimum read length
- max_length: maximum read length
- mean_length: mean read length
- median_length: median read length
- seq_lengths_dist: distribution of lengths for all
reads
- seq_lengths_masked_dist: distribution of lengths
for masked reads
- seq_lengths_padded_dist: distribution of lengths
for padded reads
The distributions are each themselves dictionaries
where the keys are read lengths and the values are
the number of reads with the matching length; note
that only lengths with a non-zero number of reads
are included (zeroes are implied for all other
lengths).
Arguments:
fastq (str): path to Fastq file
outfile (str): optional, path to output JSON file
show_progress (bool): if True then print message
to stdout every 100000 reads indicating progress
(default: operate silently)
limit (int): if set then only process this number
of reads from the head of the Fastq and return
stats based on these (default: process all reads
in the file)
Returns:
Dictionary: containing the metrics for the Fastq.
"""
nreads = 0
nreads_masked = 0
nreads_padded = 0
sequence_length = dict()
reads_all_n = dict()
reads_padded = dict()
if show_progress:
print("\n%s" % fastq)
for r in FastqIterator(fastq):
nreads += 1
if show_progress and nreads % 100000 == 0:
print("...%d reads" % nreads)
if limit and nreads == limit:
logging.info("Stopping at limit: %d" % limit)
break
seqlen = r.seqlen
try:
sequence_length[seqlen] += 1
except KeyError:
sequence_length[seqlen] = 1
nns = 0
seqlen_no_ns = len(r.sequence.strip('N'))
if seqlen_no_ns == 0:
nreads_masked += 1
try:
reads_all_n[seqlen] += 1
except KeyError:
reads_all_n[seqlen] = 1
elif seqlen_no_ns < seqlen:
nreads_padded += 1
try:
reads_padded[seqlen] += 1
except KeyError:
reads_padded[seqlen] = 1
# Get statistics
seqlens = sorted(sequence_length.keys())
min_len = min(seqlens)
max_len = max(seqlens)
mean_len = float(sum([l*sequence_length[l] for l in seqlens]))\
/float(nreads)
median_read = nreads // 2
median_len = None
read_count = 0
for l in seqlens:
read_count += sequence_length[l]
if read_count >= median_read:
median_len = l
break
frac_reads_masked = float(nreads_masked) / nreads * 100.0
frac_reads_padded = float(nreads_padded) / nreads * 100.0
# Build dictionary for output
stats = dict(fastq=fastq,
nreads=nreads,
nreads_masked=nreads_masked,
nreads_padded=nreads_padded,
frac_reads_masked=frac_reads_masked,
frac_reads_padded=frac_reads_padded,
min_length=min_len,
max_length=max_len,
mean_length=mean_len,
median_length=median_len,
seq_lengths_dist=sequence_length,
seq_lengths_masked_dist=reads_all_n,
seq_lengths_padded_dist=reads_padded)
# Output to file
if outfile:
with open(outfile, 'wt') as fp:
json.dump(stats, fp, sort_keys=True, indent=4)
# Return stats
return stats
