def test_get_n_lines_not_multiple_of_4(data):
with pytest.raises(SystemExit):
preprocessing.get_n_lines(pytest.corrupt_R1_path)
def test_get_n_lines(data):
assert preprocessing.get_n_lines(pytest.correct_R1_path) == (200 * 4)
def main():
# Create logger and stream handler
logger = logging.getLogger("cite_seq_count")
logger.setLevel(logging.CRITICAL)
ch = logging.StreamHandler()
ch.setLevel(logging.CRITICAL)
formatter = logging.Formatter(
"%(asctime)s - %(name)s - %(levelname)s - %(message)s"
)
ch.setFormatter(formatter)
logger.addHandler(ch)
start_time = time.time()
parser = get_args()
if not sys.argv[1:]:
parser.print_help(file=sys.stderr)
sys.exit(2)
# Parse arguments.
args = parser.parse_args()
if args.whitelist:
print("Loading whitelist")
(whitelist, args.bc_threshold) = preprocessing.parse_whitelist_csv(
filename=args.whitelist,
barcode_length=args.cb_last - args.cb_first + 1,
collapsing_threshold=args.bc_threshold,
)
else:
whitelist = False
# Load TAGs/ABs.
ab_map = preprocessing.parse_tags_csv(args.tags)
ab_map = preprocessing.check_tags(ab_map, args.max_error)
# Identify input file(s)
read1_paths, read2_paths = preprocessing.get_read_paths(
args.read1_path, args.read2_path
)
# preprocessing and processing occur in separate loops so the program can crash earlier if
# one of the inputs is not valid.
read1_lengths = []
read2_lengths = []
for read1_path, read2_path in zip(read1_paths, read2_paths):
# Get reads length. So far, there is no validation for Read2.
read1_lengths.append(preprocessing.get_read_length(read1_path))
read2_lengths.append(preprocessing.get_read_length(read2_path))
# Check Read1 length against CELL and UMI barcodes length.
(
barcode_slice,
umi_slice,
barcode_umi_length,
) = preprocessing.check_barcodes_lengths(
read1_lengths[-1],
args.cb_first,
args.cb_last,
args.umi_first,
args.umi_last,
)
# Ensure all files have the same input length
# if len(set(read1_lengths)) != 1:
# sys.exit('Input barcode fastqs (read1) do not all have same length.\nExiting')
# Initialize the counts dicts that will be generated from each input fastq pair
final_results = defaultdict(lambda: defaultdict(Counter))
umis_per_cell = Counter()
reads_per_cell = Counter()
merged_no_match = Counter()
number_of_samples = len(read1_paths)
n_reads = 0
# Print a statement if multiple files are run.
if number_of_samples != 1:
print("Detected {} files to run on.".format(number_of_samples))
for read1_path, read2_path in zip(read1_paths, read2_paths):
if args.first_n:
n_lines = (args.first_n * 4) / number_of_samples
else:
n_lines = preprocessing.get_n_lines(read1_path)
n_reads += int(n_lines / 4)
n_threads = args.n_threads
print("Started mapping")
print("Processing {:,} reads".format(n_reads))
# Run with one process
if n_threads <= 1 or n_reads < 1000001:
print("CITE-seq-Count is running with one core.")
(_final_results, _merged_no_match) = processing.map_reads(
read1_path=read1_path,
read2_path=read2_path,
tags=ab_map,
barcode_slice=barcode_slice,
umi_slice=umi_slice,
indexes=[0, n_reads],
whitelist=whitelist,
debug=args.debug,
start_trim=args.start_trim,
maximum_distance=args.max_error,
sliding_window=args.sliding_window,
)
print("Mapping done")
_umis_per_cell = Counter()
_reads_per_cell = Counter()
for cell_barcode, counts in _final_results.items():
_umis_per_cell[cell_barcode] = sum([len(counts[UMI]) for UMI in counts])
_reads_per_cell[cell_barcode] = sum(
[sum(counts[UMI].values()) for UMI in counts]
)
else:
# Run with multiple processes
print("CITE-seq-Count is running with {} cores.".format(n_threads))
p = Pool(processes=n_threads)
chunk_indexes = preprocessing.chunk_reads(n_reads, n_threads)
parallel_results = []
for indexes in chunk_indexes:
p.apply_async(
processing.map_reads,
args=(
read1_path,
read2_path,
ab_map,
barcode_slice,
umi_slice,
indexes,
whitelist,
args.debug,
args.start_trim,
args.max_error,
args.sliding_window,
),
callback=parallel_results.append,
error_callback=sys.stderr,
)
p.close()
p.join()
print("Mapping done")
print("Merging results")
(
_final_results,
_umis_per_cell,
_reads_per_cell,
_merged_no_match,
) = processing.merge_results(parallel_results=parallel_results)
del parallel_results
# Update the overall counts dicts
umis_per_cell.update(_umis_per_cell)
reads_per_cell.update(_reads_per_cell)
merged_no_match.update(_merged_no_match)
for cell_barcode in _final_results:
for tag in _final_results[cell_barcode]:
if tag in final_results[cell_barcode]:
# Counter + Counter = Counter
final_results[cell_barcode][tag] += _final_results[cell_barcode][
tag
]
else:
# Explicitly save the counter to that tag
final_results[cell_barcode][tag] = _final_results[cell_barcode][tag]
ordered_tags_map = OrderedDict()
for i, tag in enumerate(ab_map.values()):
ordered_tags_map[tag] = i
ordered_tags_map["unmapped"] = i + 1
# Correct cell barcodes
if args.bc_threshold > 0:
if len(umis_per_cell) <= args.expected_cells:
print(
"Number of expected cells, {}, is higher "
"than number of cells found {}.\nNot performing"
"cell barcode correction"
"".format(args.expected_cells, len(umis_per_cell))
)
bcs_corrected = 0
else:
print("Correcting cell barcodes")
if not whitelist:
(
final_results,
umis_per_cell,
bcs_corrected,
) = processing.correct_cells(
final_results=final_results,
reads_per_cell=reads_per_cell,
umis_per_cell=umis_per_cell,
expected_cells=args.expected_cells,
collapsing_threshold=args.bc_threshold,
ab_map=ordered_tags_map,
)
else:
(
final_results,
umis_per_cell,
bcs_corrected,
) = processing.correct_cells_whitelist(
final_results=final_results,
umis_per_cell=umis_per_cell,
whitelist=whitelist,
collapsing_threshold=args.bc_threshold,
ab_map=ordered_tags_map,
)
else:
bcs_corrected = 0
# If given, use whitelist for top cells
if whitelist:
top_cells = whitelist
# Add potential missing cell barcodes.
for missing_cell in whitelist:
if missing_cell in final_results:
continue
else:
final_results[missing_cell] = dict()
for TAG in ordered_tags_map:
final_results[missing_cell][TAG] = Counter()
top_cells.add(missing_cell)
else:
# Select top cells based on total umis per cell
top_cells_tuple = umis_per_cell.most_common(args.expected_cells)
top_cells = set([pair[0] for pair in top_cells_tuple])
# UMI correction
if args.no_umi_correction:
# Don't correct
umis_corrected = 0
aberrant_cells = set()
else:
# Correct UMIS
(final_results, umis_corrected, aberrant_cells) = processing.correct_umis(
final_results=final_results,
collapsing_threshold=args.umi_threshold,
top_cells=top_cells,
max_umis=20000,
)
# Remove aberrant cells from the top cells
for cell_barcode in aberrant_cells:
top_cells.remove(cell_barcode)
# Create sparse aberrant cells matrix
(umi_aberrant_matrix, read_aberrant_matrix) = processing.generate_sparse_matrices(
final_results=final_results,
ordered_tags_map=ordered_tags_map,
top_cells=aberrant_cells,
)
# Write uncorrected cells to dense output
io.write_dense(
sparse_matrix=umi_aberrant_matrix,
index=list(ordered_tags_map.keys()),
columns=aberrant_cells,
outfolder=os.path.join(args.outfolder, "uncorrected_cells"),
filename="dense_umis.tsv",
)
# Create sparse matrices for results
(umi_results_matrix, read_results_matrix) = processing.generate_sparse_matrices(
final_results=final_results,
ordered_tags_map=ordered_tags_map,
top_cells=top_cells,
)
# Write umis to file
io.write_to_files(
sparse_matrix=umi_results_matrix,
top_cells=top_cells,
ordered_tags_map=ordered_tags_map,
data_type="umi",
outfolder=args.outfolder,
)
# Write reads to file
io.write_to_files(
sparse_matrix=read_results_matrix,
top_cells=top_cells,
ordered_tags_map=ordered_tags_map,
data_type="read",
outfolder=args.outfolder,
)
# Write unmapped sequences
io.write_unmapped(
merged_no_match=merged_no_match,
top_unknowns=args.unknowns_top,
outfolder=args.outfolder,
filename=args.unmapped_file,
)
# Create report and write it to disk
create_report(
n_reads=n_reads,
reads_per_cell=reads_per_cell,
no_match=merged_no_match,
version=version,
start_time=start_time,
ordered_tags_map=ordered_tags_map,
umis_corrected=umis_corrected,
bcs_corrected=bcs_corrected,
bad_cells=aberrant_cells,
args=args,
)
# Write dense matrix to disk if requested
if args.dense:
print("Writing dense format output")
io.write_dense(
sparse_matrix=umi_results_matrix,
index=list(ordered_tags_map.keys()),
columns=top_cells,
outfolder=args.outfolder,
filename="dense_umis.tsv",
)
def main():
#Create logger and stream handler
logger = logging.getLogger('cite_seq_count')
logger.setLevel(logging.CRITICAL)
ch = logging.StreamHandler()
ch.setLevel(logging.CRITICAL)
formatter = logging.Formatter(
'%(asctime)s - %(name)s - %(levelname)s - %(message)s')
ch.setFormatter(formatter)
logger.addHandler(ch)
start_time = time.time()
parser = get_args()
if not sys.argv[1:]:
parser.print_help(file=sys.stderr)
sys.exit(2)
# Parse arguments.
args = parser.parse_args()
if args.whitelist:
(whitelist, args.bc_threshold) = preprocessing.parse_whitelist_csv(
filename=args.whitelist,
barcode_length=args.cb_last - args.cb_first + 1,
collapsing_threshold=args.bc_threshold)
else:
whitelist = False
# Load TAGs/ABs.
ab_map = preprocessing.parse_tags_csv(args.tags)
ab_map = preprocessing.check_tags(ab_map, args.max_error)
# Get reads length. So far, there is no validation for Read2.
read1_length = preprocessing.get_read_length(args.read1_path)
read2_length = preprocessing.get_read_length(args.read2_path)
# Check Read1 length against CELL and UMI barcodes length.
(barcode_slice, umi_slice,
barcode_umi_length) = preprocessing.check_barcodes_lengths(
read1_length, args.cb_first, args.cb_last, args.umi_first,
args.umi_last)
if args.first_n:
n_lines = args.first_n * 4
else:
n_lines = preprocessing.get_n_lines(args.read1_path)
n_reads = int(n_lines / 4)
n_threads = args.n_threads
print('Started mapping')
print('Processing {:,} reads'.format(n_reads))
#Run with one process
if n_threads <= 1 or n_reads < 1000001:
print('CITE-seq-Count is running with one core.')
(final_results, merged_no_match) = processing.map_reads(
read1_path=args.read1_path,
read2_path=args.read2_path,
tags=ab_map,
barcode_slice=barcode_slice,
umi_slice=umi_slice,
indexes=[0, n_reads],
whitelist=whitelist,
debug=args.debug,
start_trim=args.start_trim,
maximum_distance=args.max_error,
sliding_window=args.sliding_window)
print('Mapping done')
umis_per_cell = Counter()
reads_per_cell = Counter()
for cell_barcode, counts in final_results.items():
umis_per_cell[cell_barcode] = sum(
[len(counts[UMI]) for UMI in counts])
reads_per_cell[cell_barcode] = sum(
[sum(counts[UMI].values()) for UMI in counts])
else:
# Run with multiple processes
print('CITE-seq-Count is running with {} cores.'.format(n_threads))
p = Pool(processes=n_threads)
chunk_indexes = preprocessing.chunk_reads(n_reads, n_threads)
parallel_results = []
for indexes in chunk_indexes:
p.apply_async(processing.map_reads,
args=(args.read1_path, args.read2_path, ab_map,
barcode_slice, umi_slice, indexes, whitelist,
args.debug, args.start_trim, args.max_error,
args.sliding_window),
callback=parallel_results.append,
error_callback=sys.stderr)
p.close()
p.join()
print('Mapping done')
print('Merging results')
(final_results, umis_per_cell, reads_per_cell,
merged_no_match) = processing.merge_results(
parallel_results=parallel_results)
del (parallel_results)
ordered_tags_map = OrderedDict()
for i, tag in enumerate(ab_map.values()):
ordered_tags_map[tag] = i
ordered_tags_map['unmapped'] = i + 1
# Correct cell barcodes
if (len(umis_per_cell) <= args.expected_cells):
print("Number of expected cells, {}, is higher " \
"than number of cells found {}.\nNot performing" \
"cell barcode correction" \
"".format(args.expected_cells, len(umis_per_cell)))
bcs_corrected = 0
else:
print('Correcting cell barcodes')
if not whitelist:
(final_results, umis_per_cell,
bcs_corrected) = processing.correct_cells(
final_results=final_results,
umis_per_cell=umis_per_cell,
expected_cells=args.expected_cells,
collapsing_threshold=args.bc_threshold)
else:
(final_results, umis_per_cell,
bcs_corrected) = processing.correct_cells_whitelist(
final_results=final_results,
umis_per_cell=umis_per_cell,
whitelist=whitelist,
collapsing_threshold=args.bc_threshold)
# Correct umi barcodes
if not whitelist:
top_cells_tuple = umis_per_cell.most_common(args.expected_cells)
top_cells = set([pair[0] for pair in top_cells_tuple])
# Sort cells by number of mapped umis
else:
top_cells = whitelist
# Add potential missing cell barcodes.
for missing_cell in whitelist:
if missing_cell in final_results:
continue
else:
final_results[missing_cell] = dict()
for TAG in ordered_tags_map:
final_results[missing_cell][TAG] = Counter()
top_cells.add(missing_cell)
#If we want umi correction
if not args.no_umi_correction:
(final_results, umis_corrected,
aberrant_cells) = processing.correct_umis(
final_results=final_results,
collapsing_threshold=args.umi_threshold,
top_cells=top_cells,
max_umis=20000)
else:
umis_corrected = 0
aberrant_cells = set()
for cell_barcode in aberrant_cells:
top_cells.remove(cell_barcode)
#Create sparse aberrant cells matrix
(umi_aberrant_matrix,
read_aberrant_matrix) = processing.generate_sparse_matrices(
final_results=final_results,
ordered_tags_map=ordered_tags_map,
top_cells=aberrant_cells)
#Write uncorrected cells to dense output
io.write_dense(sparse_matrix=umi_aberrant_matrix,
index=list(ordered_tags_map.keys()),
columns=aberrant_cells,
outfolder=os.path.join(args.outfolder, 'uncorrected_cells'),
filename='dense_umis.tsv')
(umi_results_matrix,
read_results_matrix) = processing.generate_sparse_matrices(
final_results=final_results,
ordered_tags_map=ordered_tags_map,
top_cells=top_cells)
# Write umis to file
io.write_to_files(sparse_matrix=umi_results_matrix,
top_cells=top_cells,
ordered_tags_map=ordered_tags_map,
data_type='umi',
outfolder=args.outfolder)
# Write reads to file
io.write_to_files(sparse_matrix=read_results_matrix,
top_cells=top_cells,
ordered_tags_map=ordered_tags_map,
data_type='read',
outfolder=args.outfolder)
top_unmapped = merged_no_match.most_common(args.unknowns_top)
with open(os.path.join(args.outfolder, args.unmapped_file),
'w') as unknown_file:
unknown_file.write('tag,count\n')
for element in top_unmapped:
unknown_file.write('{},{}\n'.format(element[0], element[1]))
create_report(n_reads=n_reads,
reads_per_cell=reads_per_cell,
no_match=merged_no_match,
version=version,
start_time=start_time,
ordered_tags_map=ordered_tags_map,
umis_corrected=umis_corrected,
bcs_corrected=bcs_corrected,
bad_cells=aberrant_cells,
args=args)
if args.dense:
print('Writing dense format output')
io.write_dense(sparse_matrix=umi_results_matrix,
index=list(ordered_tags_map.keys()),
columns=top_cells,
outfolder=args.outfolder,
filename='dense_umis.tsv')
def main():
start_time = time.time()
parser = get_args()
if not sys.argv[1:]:
parser.print_help(file=sys.stderr)
sys.exit(2)
# Parse arguments.
args = parser.parse_args()
if args.whitelist:
whitelist = preprocessing.parse_whitelist_csv(
args.whitelist, args.cb_last - args.cb_first + 1)
else:
whitelist = None
# Load TAGs/ABs.
ab_map = preprocessing.parse_tags_csv(args.tags)
ab_map = preprocessing.check_tags(ab_map, args.max_error)
# Get reads length. So far, there is no validation for Read2.
read1_length = preprocessing.get_read_length(args.read1_path)
read2_length = preprocessing.get_read_length(args.read2_path)
# Check Read1 length against CELL and UMI barcodes length.
(barcode_slice, umi_slice,
barcode_umi_length) = preprocessing.check_barcodes_lengths(
read1_length, args.cb_first, args.cb_last, args.umi_first,
args.umi_last)
if args.first_n:
n_lines = args.first_n * 4
else:
n_lines = preprocessing.get_n_lines(args.read1_path)
n_reads = int(n_lines / 4)
n_threads = args.n_threads
print('Started mapping')
#Run with one process
if n_threads <= 1 or n_reads < 1000001:
print('CITE-seq-Count is running with one core.')
(final_results, merged_no_match) = processing.map_reads(
read1_path=args.read1_path,
read2_path=args.read2_path,
tags=ab_map,
barcode_slice=barcode_slice,
umi_slice=umi_slice,
indexes=[0, n_reads],
whitelist=whitelist,
debug=args.debug,
start_trim=args.start_trim,
maximum_distance=args.max_error)
print('Mapping done')
umis_per_cell = Counter()
reads_per_cell = Counter()
for cell_barcode, counts in final_results.items():
umis_per_cell[cell_barcode] = sum(
[len(counts[UMI]) for UMI in counts if UMI != 'unmapped'])
reads_per_cell[cell_barcode] = sum([
sum(counts[UMI].values()) for UMI in counts
if UMI != 'unmapped'
])
else:
# Run with multiple processes
print('CITE-seq-Count is running with {} cores.'.format(n_threads))
p = Pool(processes=n_threads)
chunk_indexes = preprocessing.chunk_reads(n_reads, n_threads)
parallel_results = []
for indexes in chunk_indexes:
p.apply_async(processing.map_reads,
args=(args.read1_path, args.read2_path, ab_map,
barcode_slice, umi_slice, indexes, whitelist,
args.debug, args.start_trim, args.max_error),
callback=parallel_results.append,
error_callback=sys.stderr)
p.close()
p.join()
print('Mapping done')
print('Merging results')
(final_results, umis_per_cell, reads_per_cell,
merged_no_match) = processing.merge_results(
parallel_results=parallel_results)
del (parallel_results)
# Correct cell barcodes
(final_results, umis_per_cell, bcs_corrected) = processing.correct_cells(
final_results=final_results,
umis_per_cell=umis_per_cell,
expected_cells=args.expected_cells,
collapsing_threshold=args.bc_threshold)
# Correct umi barcodes
(final_results, umis_corrected) = processing.correct_umis(
final_results=final_results, collapsing_threshold=args.umi_threshold)
ordered_tags_map = OrderedDict()
for i, tag in enumerate(ab_map.values()):
ordered_tags_map[tag] = i
ordered_tags_map['unmapped'] = i + 1
# Sort cells by number of mapped umis
if not whitelist:
top_cells_tuple = umis_per_cell.most_common(args.expected_cells)
top_cells = set([pair[0] for pair in top_cells_tuple])
else:
top_cells = whitelist
# Add potential missing cell barcodes.
for missing_cell in whitelist:
if missing_cell in final_results:
continue
else:
final_results[missing_cell] = dict()
for TAG in ordered_tags_map:
final_results[missing_cell][TAG] = 0
top_cells.add(missing_cell)
(umi_results_matrix,
read_results_matrix) = processing.generate_sparse_matrices(
final_results=final_results,
ordered_tags_map=ordered_tags_map,
top_cells=top_cells)
io.write_to_files(sparse_matrix=umi_results_matrix,
top_cells=top_cells,
ordered_tags_map=ordered_tags_map,
data_type='umi',
outfolder=args.outfolder)
io.write_to_files(sparse_matrix=read_results_matrix,
top_cells=top_cells,
ordered_tags_map=ordered_tags_map,
data_type='read',
outfolder=args.outfolder)
top_unmapped = merged_no_match.most_common(args.unknowns_top)
with open(os.path.join(args.outfolder, args.unmapped_file),
'w') as unknown_file:
unknown_file.write('tag,count\n')
for element in top_unmapped:
unknown_file.write('{},{}\n'.format(element[0], element[1]))
create_report(n_reads=n_reads,
reads_per_cell=reads_per_cell,
no_match=merged_no_match,
version=version,
start_time=start_time,
ordered_tags_map=ordered_tags_map,
umis_corrected=umis_corrected,
bcs_corrected=bcs_corrected,
args=args)
if args.dense:
print('Writing dense format output')
io.write_dense(sparse_matrix=umi_results_matrix,
index=list(ordered_tags_map.keys()),
columns=top_cells,
file_path=os.path.join(args.outfolder,
'dense_umis.tsv'))
