def alignAcross(data, field_map, muscle_exec=default_muscle_exec):
"""
Multiple aligns sequence fields column wise
Arguments:
data : DbData object with Receptor objects to process.
field_map : a dictionary of {input sequence : output sequence) field names to multiple align.
muscle_exec : the MUSCLE executable.
Returns:
changeo.Multiprocessing.DbResult : object containing Receptor objects with multiple aligned sequence fields.
"""
# Define return object
result = DbResult(data.id, data.data)
result.results = data.data
result.valid = True
# Fail invalid groups
if result.id is None:
result.log = None
result.valid = False
return result
seq_fields = list(field_map.keys())
for f in seq_fields:
seq_list = [
SeqRecord(r.getSeq(f), id=r.sequence_id.replace(' ', '_'))
for r in data.data
]
seq_aln = runMuscle(seq_list, aligner_exec=muscle_exec)
if seq_aln is not None:
aln_map = {x.id: i for i, x in enumerate(seq_aln)}
for i, r in enumerate(result.results, start=1):
idx = aln_map[r.sequence_id.replace(' ', '_')]
seq = str(seq_aln[idx].seq)
r.annotations[field_map[f]] = seq
result.log['%s-%s' % (f, r.sequence_id)] = seq
else:
result.valid = False
#for r in result.results: print r.annotations
return result
def alignBlocks(data, seq_fields, muscle_exec=default_muscle_exec):
"""
Multiple aligns blocks of sequence fields together
Arguments:
data : a DbData object with IgRecords to process.
seq_fields : the sequence fields to multiple align.
muscle_exec : the MUSCLE executable.
Returns:
changeo.Multiprocessing.DbResult : object containing IgRecords with multiple aligned sequence fields.
"""
# Define return object
result = DbResult(data.id, data.data)
result.results = data.data
result.valid = True
# Fail invalid groups
if result.id is None:
result.log = None
result.valid = False
return result
seq_list = [SeqRecord(r.getSeqField(f), id='%s_%s' % (r.id, f)) for f in seq_fields \
for r in data.data]
seq_aln = runMuscle(seq_list, aligner_exec=muscle_exec)
if seq_aln is not None:
aln_map = {x.id: i for i, x in enumerate(seq_aln)}
for i, r in enumerate(result.results, start=1):
for f in seq_fields:
idx = aln_map['%s_%s' % (r.id, f)]
seq = str(seq_aln[idx].seq)
r.annotations['%s_ALIGN' % f] = seq
result.log['%s-%s' % (f, r.id)] = seq
else:
result.valid = False
#for r in result.results: print r.annotations
return result
def alignWithin(data, field_map, muscle_exec=default_muscle_exec):
"""
Multiple aligns sequence fields within a row
Arguments:
data : DbData object with Receptor objects to process.
field_map : a dictionary of {input sequence : output sequence) field names to multiple align.
muscle_exec : the MUSCLE executable.
Returns:
changeo.Multiprocessing.DbResult : object containing Receptor objects with multiple aligned sequence fields.
"""
# Define return object
result = DbResult(data.id, data.data)
result.results = data.data
result.valid = True
# Fail invalid groups
if result.id is None:
result.log = None
result.valid = False
return result
record = data.data
seq_fields = list(field_map.keys())
seq_list = [SeqRecord(record.getSeq(f), id=f) for f in seq_fields]
seq_aln = runMuscle(seq_list, aligner_exec=muscle_exec)
if seq_aln is not None:
aln_map = {x.id: i for i, x in enumerate(seq_aln)}
for f in seq_fields:
idx = aln_map[f]
seq = str(seq_aln[idx].seq)
record.annotations[field_map[f]] = seq
result.log[f] = seq
else:
result.valid = False
return result
def alignWithin(data, seq_fields, muscle_exec=default_muscle_exec):
"""
Multiple aligns sequence fields within a row
Arguments:
data : a DbData object with an IgRecords to process.
seq_fields : the sequence fields to multiple align.
muscle_exec : the MUSCLE executable.
Returns:
changeo.Multiprocessing.DbResult : object containing IgRecords with multiple aligned sequence fields.
"""
# Define return object
result = DbResult(data.id, data.data)
result.results = data.data
result.valid = True
# Fail invalid groups
if result.id is None:
result.log = None
result.valid = False
return result
record = data.data
seq_list = [SeqRecord(record.getSeqField(f), id=f) for f in seq_fields]
seq_aln = runMuscle(seq_list, aligner_exec=muscle_exec)
if seq_aln is not None:
aln_map = {x.id: i for i, x in enumerate(seq_aln)}
for f in seq_fields:
idx = aln_map[f]
seq = str(seq_aln[idx].seq)
record.annotations['%s_ALIGN' % f] = seq
result.log[f] = seq
else:
result.valid = False
return result
def processQueueClust(alive, data_queue, result_queue, clone_func, clone_args):
"""
Pulls from data queue, performs calculations, and feeds results queue
Arguments:
alive = a multiprocessing.Value boolean controlling whether processing continues
if False exit process
data_queue = a multiprocessing.Queue holding data to process
result_queue = a multiprocessing.Queue to hold processed results
clone_func = the function to call for calculating pairwise distances between sequences
clone_args = a dictionary of arguments to pass to clone_func
Returns:
None
"""
try:
# print 'START WORK', alive.value
# Iterator over data queue until sentinel object reached
while alive.value:
# Get data from queue
if data_queue.empty(): continue
else: data = data_queue.get()
# Exit upon reaching sentinel
if data is None: break
# print "WORK", alive.value, data['id']
# Define result object for iteration and get data records
records = data.data
result = DbResult(data.id, records)
# Create row of distance matrix and check for error
dist_row = clone_func(records, **clone_args) if data else None
if dist_row is not None:
result.results = dist_row
result.valid = True
# Feed results to result queue
result_queue.put(result)
else:
sys.stderr.write('PID %s: Error in sibling process detected. Cleaning up.\n' \
% os.getpid())
return None
except:
#sys.stderr.write('Exception in worker\n')
alive.value = False
raise
return None
def processQueueClust(alive, data_queue, result_queue, clone_func, clone_args):
"""
Pulls from data queue, performs calculations, and feeds results queue
Arguments:
alive = a multiprocessing.Value boolean controlling whether processing continues
if False exit process
data_queue = a multiprocessing.Queue holding data to process
result_queue = a multiprocessing.Queue to hold processed results
clone_func = the function to call for calculating pairwise distances between sequences
clone_args = a dictionary of arguments to pass to clone_func
Returns:
None
"""
try:
# print 'START WORK', alive.value
# Iterator over data queue until sentinel object reached
while alive.value:
# Get data from queue
if data_queue.empty(): continue
else: data = data_queue.get()
# Exit upon reaching sentinel
if data is None: break
# print "WORK", alive.value, data['id']
# Define result object for iteration and get data records
records = data.data
result = DbResult(data.id, records)
# Create row of distance matrix and check for error
dist_row = clone_func(records, **clone_args) if data else None
if dist_row is not None:
result.results = dist_row
result.valid = True
# Feed results to result queue
result_queue.put(result)
else:
sys.stderr.write('PID %s: Error in sibling process detected. Cleaning up.\n' \
% os.getpid())
return None
except:
#sys.stderr.write('Exception in worker\n')
alive.value = False
raise
return None
def processQueue(alive, data_queue, result_queue, clone_func, clone_args):
"""
Pulls from data queue, performs calculations, and feeds results queue
Arguments:
alive = a multiprocessing.Value boolean controlling whether processing continues
if False exit process
data_queue = a multiprocessing.Queue holding data to process
result_queue = a multiprocessing.Queue to hold processed results
clone_func = the function to call for clonal assignment
clone_args = a dictionary of arguments to pass to clone_func
Returns:
None
"""
try:
# Iterator over data queue until sentinel object reached
while alive.value:
# Get data from queue
if data_queue.empty(): continue
else: data = data_queue.get()
# Exit upon reaching sentinel
if data is None: break
# Define result object for iteration and get data records
records = data.data
result = DbResult(data.id, records)
# Check for invalid data (due to failed indexing) and add failed result
if not data:
result_queue.put(result)
continue
# Add V(D)J to log
result.log['ID'] = ','.join([str(x) for x in data.id])
result.log['VALLELE'] = ','.join(set([(r.getVAllele() or '') for r in records]))
result.log['DALLELE'] = ','.join(set([(r.getDAllele() or '') for r in records]))
result.log['JALLELE'] = ','.join(set([(r.getJAllele() or '') for r in records]))
result.log['JUNCLEN'] = ','.join(set([(str(len(r.junction)) or '0') for r in records]))
result.log['SEQUENCES'] = len(records)
# Checking for preclone failure and assign clones
clones = clone_func(records, **clone_args) if data else None
# import cProfile
# prof = cProfile.Profile()
# clones = prof.runcall(clone_func, records, **clone_args)
# prof.dump_stats('worker-%d.prof' % os.getpid())
if clones is not None:
result.results = clones
result.valid = True
result.log['CLONES'] = len(clones)
else:
result.log['CLONES'] = 0
# Feed results to result queue
result_queue.put(result)
else:
sys.stderr.write('PID %s: Error in sibling process detected. Cleaning up.\n' \
% os.getpid())
return None
except:
#sys.stderr.write('Exception in worker\n')
alive.value = False
raise
return None
def processQueue(alive,
data_queue,
result_queue,
max_missing=default_max_missing,
clone_func=distanceClones,
clone_args={}):
"""
Pulls from data queue, performs calculations, and feeds results queue
Arguments:
alive : a multiprocessing.Value boolean controlling whether processing continues
if False exit process
data_queue : a multiprocessing.Queue holding data to process
result_queue : a multiprocessing.Queue to hold processed results
max_missing : maximum number of non-ACGT characters to allow in the junction sequence.
clone_func : the function to call for clonal assignment
clone_args : a dictionary of arguments to pass to clone_func
Returns:
None
"""
try:
# Iterator over data queue until sentinel object reached
while alive.value:
# Get data from queue
if data_queue.empty(): continue
else: data = data_queue.get()
# Exit upon reaching sentinel
if data is None: break
# Define result object for iteration and get data records
result = DbResult(data.id, data.data)
# Check for invalid data (due to failed indexing) and add failed result
if not data:
result_queue.put(result)
continue
# Filter records based on missing content
seq_field = clone_args[
'seq_field'] if 'seq_field' in clone_args else 'JUNCTION'
filtered = filterMissing(data.data,
field=seq_field,
max_missing=max_missing)
records = filtered['pass']
result.failed = filtered['fail']
# Add V(D)J to log
result.log['ID'] = ','.join([str(x) for x in data.id])
result.log['VALLELE'] = ','.join(
set([(r.getVAllele() or '') for r in data.data]))
result.log['DALLELE'] = ','.join(
set([(r.getDAllele() or '') for r in data.data]))
result.log['JALLELE'] = ','.join(
set([(r.getJAllele() or '') for r in data.data]))
result.log['JUNCLEN'] = ','.join(
set([(str(len(r.junction)) or '0') for r in data.data]))
result.log['PASSCOUNT'] = len(records)
result.log['FAILCOUNT'] = len(result.failed)
# Checking for preclone failure and assign clones
clones = clone_func(records, **clone_args) if records else None
# import cProfile
# prof = cProfile.Profile()
# clones = prof.runcall(clone_func, records, **clone_args)
# prof.dump_stats('worker-%d.prof' % os.getpid())
if clones is not None:
result.results = clones
result.valid = True
result.log['CLONES'] = len(clones)
else:
result.log['CLONES'] = 0
# Feed results to result queue
result_queue.put(result)
else:
sys.stderr.write('PID %s: Error in sibling process detected. Cleaning up.\n' \
% os.getpid())
return None
except:
#sys.stderr.write('Exception in worker\n')
alive.value = False
raise
return None