def count_reads_in_features(
sam_filename,
gff_filename,
order,
max_buffer_size,
stranded,
overlap_mode,
multimapped_mode,
secondary_alignment_mode,
supplementary_alignment_mode,
feature_type,
id_attribute,
additional_attributes,
quiet,
minaqual,
samout,
samout_format,
output_delimiter,
output_filename,
cb_tag,
ub_tag,
):
'''Count reads in features, parallelizing by file'''
if samout is not None:
# Try to open samout file early in case any of them has issues
if samout_format in ('SAM', 'sam'):
with open(samout, 'w'):
pass
else:
# We don't have a template if the input is stdin
if sam_filename != '-':
with pysam.AlignmentFile(sam_filename, 'r') as sf:
with pysam.AlignmentFile(samout, 'w', template=sf):
pass
# Try to open samfiles to fail early in case any of them is not there
if sam_filename != '-':
with pysam.AlignmentFile(sam_filename, 'r') as sf:
pass
# Prepare features
gff = HTSeq.GFF_Reader(gff_filename)
feature_scan = HTSeq.make_feature_genomicarrayofsets(
gff,
id_attribute,
feature_type=feature_type,
additional_attributes=additional_attributes,
stranded=stranded != 'no',
verbose=not quiet,
)
features = feature_scan['features']
attributes = feature_scan['attributes']
feature_attr = sorted(attributes.keys())
if len(feature_attr) == 0:
sys.stderr.write("Warning: No features of type '%s' found.\n" %
feature_type)
# Count reads
results = count_reads_with_barcodes(
sam_filename,
features,
feature_attr,
order,
max_buffer_size,
stranded,
overlap_mode,
multimapped_mode,
secondary_alignment_mode,
supplementary_alignment_mode,
feature_type,
id_attribute,
additional_attributes,
quiet,
minaqual,
samout_format,
samout,
cb_tag,
ub_tag,
)
# Cell barcodes
cbs = results['cell_barcodes']
counts = results['counts']
# Write output
other_features = [
'__no_feature',
'__ambiguous',
'__too_low_aQual',
'__not_aligned',
'__alignment_not_unique',
]
pad = ['' for attr in additional_attributes]
# Header
fields = [''] + pad + cbs
line = output_delimiter.join(fields)
if output_filename == '':
print(line)
else:
with open(output_filename, 'w') as f:
f.write(line)
f.write('\n')
# Features
for ifn, fn in enumerate(feature_attr):
fields = [fn] + attributes[fn] + [str(counts[cb][fn]) for cb in cbs]
line = output_delimiter.join(fields)
if output_filename == '':
print(line)
else:
with open(output_filename, 'a') as f:
f.write(line)
f.write('\n')
# Other features
for fn in other_features:
fields = [fn] + pad + [str(counts[cb][fn]) for cb in cbs]
line = output_delimiter.join(fields)
if output_filename == '':
print(line)
else:
with open(output_filename, 'a') as f:
f.write(line)
f.write('\n')
def count_reads_in_features(args):
"""Count reads in features, parallelizing by file
Args:
args: ArgumentParser args, i.e. each argument is a property of this
instance. Check the CLI parsing function below for a full list
of properties, i.e. command-line options.
This function can be conceptually split into the following steps:
1. Load features from GTF file into memory
2. Parse the reads from each BAM file in parallel or series
3. Write output table
Step 2 can be further split into two main components:
1. Find what features overlap with each read/read pair
2. Assign that read/pair to a feature (if unique) or a corner case
(e.g. multimappers)
"""
# Load feature GenomicArrayOfSets to mark overlaps
gff = HTSeq.GFF_Reader(args.featuresfilename)
feature_scan = HTSeq.make_feature_genomicarrayofsets(
gff,
args.idattr,
feature_type=args.feature_type,
feature_query=args.feature_query,
additional_attributes=args.additional_attributes,
stranded=args.stranded != "no",
verbose=not args.quiet,
add_chromosome_info=args.add_chromosome_info,
)
features = feature_scan["features"]
attributes = feature_scan["attributes"]
feature_attr = sorted(attributes.keys())
if len(feature_attr) == 0:
sys.stderr.write(
"Warning: No features of type '{args.feature_type}' found.\n")
# Count reads in parallel or in series
count_args, attributes = _prepare_args_for_counting(
features,
feature_attr,
attributes,
args.add_chromosome_info,
args.additional_attributes,
args.feature_query,
args.feature_type,
args.featuresfilename,
args.idattr,
args.max_buffer_size,
args.minaqual,
args.nonunique,
args.order,
args.mode,
args.quiet,
args.samfilenames,
args.samout_format,
args.samouts,
args.secondary_alignments,
args.stranded,
args.supplementary_alignments,
)
if args.nprocesses > 1:
with multiprocessing.Pool(args.nprocesses) as pool:
results = pool.starmap(count_reads_single_file, count_args)
results.sort(key=operator.itemgetter("isam"))
else:
results = list(itertools.starmap(count_reads_single_file, count_args))
# Merge and write output
_write_output(
results,
args.samfilenames,
attributes,
args.additional_attributes,
args.output_filename,
args.output_delimiter,
args.output_append,
sparse=args.counts_output_sparse,
dtype=np.float32,
)
def count_reads_in_features(
sam_filenames,
gff_filename,
order,
max_buffer_size,
stranded,
overlap_mode,
multimapped_mode,
secondary_alignment_mode,
supplementary_alignment_mode,
feature_type,
id_attribute,
additional_attributes,
quiet,
minaqual,
samouts,
samout_format,
output_delimiter,
output_filename,
output_append,
nprocesses,
feature_query,
):
'''Count reads in features, parallelizing by file'''
# Never use more CPUs than files
nprocesses = min(nprocesses, len(sam_filenames))
if samouts != []:
if len(samouts) != len(sam_filenames):
raise ValueError(
'Select the same number of input and output files')
# Try to open samout files early in case any of them has issues
if samout_format in ('SAM', 'sam'):
for samout in samouts:
with open(samout, 'w'):
pass
else:
# We don't have a template if the input is stdin
if (len(sam_filenames) != 1) or (sam_filenames[0] != '-'):
for sam_filename, samout in zip(sam_filenames, samouts):
with pysam.AlignmentFile(sam_filename, 'r') as sf:
with pysam.AlignmentFile(samout, 'w', template=sf):
pass
else:
samouts = [None for x in sam_filenames]
# Try to open samfiles to fail early in case any of them is not there
if (len(sam_filenames) != 1) or (sam_filenames[0] != '-'):
for sam_filename in sam_filenames:
with pysam.AlignmentFile(sam_filename, 'r') as sf:
pass
# Prepare features
gff = HTSeq.GFF_Reader(gff_filename)
feature_scan = HTSeq.make_feature_genomicarrayofsets(
gff,
id_attribute,
feature_type=feature_type,
feature_query=feature_query,
additional_attributes=additional_attributes,
stranded=stranded != 'no',
verbose=not quiet,
)
features = feature_scan['features']
attributes = feature_scan['attributes']
feature_attr = sorted(attributes.keys())
if len(feature_attr) == 0:
sys.stderr.write("Warning: No features of type '%s' found.\n" %
feature_type)
# Prepare arguments for counting function
args = []
for isam, (sam_filename,
samout_filename) in enumerate(zip(sam_filenames, samouts)):
args.append((
isam,
sam_filename,
features,
feature_attr,
order,
max_buffer_size,
stranded,
overlap_mode,
multimapped_mode,
secondary_alignment_mode,
supplementary_alignment_mode,
feature_type,
id_attribute,
additional_attributes,
quiet,
minaqual,
samout_format,
samout_filename,
))
# Count reads
if nprocesses > 1:
with multiprocessing.Pool(nprocesses) as pool:
results = pool.starmap(count_reads_single_file, args)
results.sort(key=operator.itemgetter('isam'))
else:
results = list(itertools.starmap(count_reads_single_file, args))
# Write output
other_features = [
('__no_feature', 'empty'),
('__ambiguous', 'ambiguous'),
('__too_low_aQual', 'lowqual'),
('__not_aligned', 'notaligned'),
('__alignment_not_unique', 'nonunique'),
]
pad = ['' for attr in additional_attributes]
for ifn, fn in enumerate(feature_attr):
fields = [fn
] + attributes[fn] + [str(r['counts'][fn]) for r in results]
line = output_delimiter.join(fields)
if output_filename == '':
print(line)
else:
omode = 'a' if output_append or (ifn > 0) else 'w'
with open(output_filename, omode) as f:
f.write(line)
f.write('\n')
for title, fn in other_features:
fields = [title] + pad + [str(r[fn]) for r in results]
line = output_delimiter.join(fields)
if output_filename == '':
print(line)
else:
with open(output_filename, 'a') as f:
f.write(line)
f.write('\n')
