def ec2emase(file_in, file_out, target_file=None):
ec = ec_file.parse(file_in)
new_shape = (len(ec._targets_list), len(ec._haplotypes_list), len(ec._ec_list))
LOG.info('Creating APM...')
LOG.debug('Shape={}'.format(new_shape))
apm = APM(shape=new_shape, haplotype_names=ec._haplotypes_list, locus_names=ec._targets_list, read_names=ec._ec_list)
LOG.debug('ec._haplotypes_list={}'.format(str(ec._haplotypes_list)))
LOG.debug('ec._targets_list[0:10]={}'.format(str(ec._targets_list[0:10])))
LOG.debug('ec._ec_list[0:10]={}'.format(str(ec._ec_list[0:10])))
# counts -> the number of times this equivalence class has appeared
apm.count = ec._ec_counts_list
counter = 0
num_haplotypes = len(ec._haplotypes_list)
try:
for alignment in ec._alignments:
#LOG.verbose(str(alignment))
ec_index = alignment[0]
target_index = alignment[1]
temp_bits = alignment[2]
if temp_bits == 0:
continue
bits = int_to_list(temp_bits, num_haplotypes)
for i, bit in enumerate(bits):
if bit:
# lid, hid, rid, value
apm.set_value(target_index, i, ec_index, 1)
except Exception, e:
_show_error()
raise e
def convert(file_in, file_out, target_file=None, emase=False):
"""
:param file_in: Input BAM/SAM file.
:param file_out: Output file name.
:param target_file: The target file is a list of main targets that will be used as main targets,
not to limit the main targets. Useful for comparison purposes between BAM files.
:param emase: Emase output or normal.
:return:
"""
LOG.info('Input File: {}'.format(file_in))
LOG.info('Output File: {}'.format(file_out))
if target_file:
LOG.info('Target File: {}'.format(target_file))
if emase:
LOG.info('Emase format requested')
main_targets = OrderedDict()
if target_file:
main_targets = parse_target_file(target_file)
if len(main_targets) == 0:
LOG.error("Unable to parse target file")
sys.exit(-1)
# ec = equivalence class
# the KEY is a comma separated string of tids
# the VALUE is the number of times this equivalence class has appeared
ec = OrderedDict()
# ec_idx = lookup to ec
# the KEY is a comma separated string of tids
# the VALUE is a number specifying the insertion order of the KEY value in ec
ec_idx = {}
# all the haplotypes
haplotypes = set()
# a lookup of tids to main_targets (Ensembl IDs)
target_idx_to_main_target = {}
# unique number of tids encountered and the count
unique_tids = {}
# unique reads
unique_reads = {}
# times encountering new read id
read_id_switch_counter = 0
same_read_target_counter = 0
try:
sam_file = pysam.Samfile(file_in, 'rb')
if len(sam_file.header) == 0:
raise Exception("BAM File has no header information")
except:
sam_file = pysam.Samfile(file_in, 'r')
if len(sam_file.header) == 0:
raise Exception("SAM File has no header information")
line_no = 0
ec_key = None
tid = None
target_ids = []
try:
read_id = None
while True:
alignment = sam_file.next()
line_no += 1
# reference_sequence_name = Column 3 from file, the Reference NAME (EnsemblID_Haplotype)
# tid = the target id, which is 0 or a positive integer mapping to entries
# within the sequence dictionary in the header section of a BAM file
# main_target = the Ensembl id of the transcript
if alignment.flag == 4:
continue
reference_sequence_name = sam_file.getrname(alignment.tid)
tid = str(alignment.tid)
main_target = reference_sequence_name.split('_')[0]
try:
unique_tids[tid] += 1
except KeyError:
unique_tids[tid] = 1
#LOG.verbose("{}\t{}\t{}".format(main_target, reference_sequence_name, tid))
if target_file:
if main_target not in main_targets:
LOG.error("Unexpected target found in BAM file: {}".format(main_target))
sys.exit(-1)
else:
if main_target not in main_targets:
main_targets[main_target] = len(main_targets)
target_idx_to_main_target[tid] = main_target
try:
haplotypes.add(reference_sequence_name.split('_')[1])
except:
LOG.info('Unable to parse Haplotype from {}'.format(reference_sequence_name))
return
# read_id = Column 1 from file, the Query template NAME
if read_id is None:
read_id = alignment.qname
try:
unique_reads[read_id] += 1
except KeyError:
unique_reads[read_id] = 1
if read_id != alignment.qname:
ec_key = ','.join(sorted(target_ids))
try:
ec[ec_key] += 1
except KeyError:
ec[ec_key] = 1
ec_idx[ec_key] = len(ec_idx)
read_id = alignment.qname
target_ids = [tid]
read_id_switch_counter += 1
else:
if tid not in target_ids:
target_ids.append(tid)
else:
same_read_target_counter += 1
if line_no % 1000000 == 0:
LOG.info("{0:,} alignments processed, with {1:,} equivalence classes".format(line_no, len(ec)))
except StopIteration:
LOG.info("{0:,} alignments processed, with {1:,} equivalence classes".format(line_no, len(ec)))
if tid not in target_ids:
target_ids.append(tid)
else:
same_read_target_counter += 1
ec_key = ','.join(sorted(target_ids))
try:
ec[ec_key] += 1
except KeyError:
ec[ec_key] = 1
ec_idx[ec_key] = len(ec_idx)
haplotypes = sorted(list(haplotypes))
LOG.info("# Unique Reads: {:,}".format(len(unique_reads)))
LOG.info("# Reads/Target Duplications: {:,}".format(same_read_target_counter))
LOG.info("# Main Targets: {:,}".format(len(main_targets)))
LOG.info("# Haplotypes: {:,}".format(len(haplotypes)))
LOG.info("# Unique Targets: {:,}".format(len(unique_tids)))
LOG.info("# Equivalence Classes: {:,}".format(len(ec)))
try:
os.remove(file_out)
except OSError:
pass
if emase:
try:
LOG.info('Creating APM...')
if LOG.isEnabledFor(VERBOSE_LEVELV_NUM):
LOG.verbose("HAPLOTYPES")
for h in haplotypes:
LOG.verbose(h)
LOG.verbose("MAIN TARGETS")
for m in main_targets:
LOG.verbose(m)
new_shape = (len(main_targets), len(haplotypes), len(ec))
ec_ids = [x for x in xrange(0, len(ec))]
LOG.debug('Shape={}'.format(new_shape))
apm = APM(shape=new_shape, haplotype_names=haplotypes, locus_names=main_targets.keys(), read_names=ec_ids)
# ec.values -> the number of times this equivalence class has appeared
apm.count = ec.values()
# k = comma seperated string of tids
# v = the count
for k, v in ec.iteritems():
arr_target_idx = k.split(",")
# get the main targets by name
temp_main_targets = set()
for idx in arr_target_idx:
temp_main_targets.add(target_idx_to_main_target[idx])
# loop through the targets and haplotypes to get the bits
for main_target in temp_main_targets:
# main_target is not an index, but a value like 'ENMUST..001'
for i, hap in enumerate(haplotypes):
read_transcript = '{}_{}'.format(main_target, hap) # now 'ENMUST..001_A'
# get the numerical tid corresponding to read_transcript
read_transcript_idx = str(sam_file.gettid(read_transcript))
if read_transcript_idx in arr_target_idx:
LOG.debug("{}\t{}\t{}".format(ec_idx[k], main_targets[main_target], i))
# main_targets[main_target] = idx of main target
# i = the haplotype
# ec_idx[k] = index of ec
apm.set_value(main_targets[main_target], i, ec_idx[k], 1)
LOG.info("Finalizing...")
apm.finalize()
apm.save(file_out, title='bam2ec')
except:
_show_error()
else:
try:
LOG.info("Generating BIN file...")
f = open(file_out, "wb")
# version
f.write(pack('<i', 1))
LOG.verbose("1\t# VERSION")
# targets
LOG.verbose("{:,}\t# NUMBER OF TARGETS".format(len(main_targets)))
f.write(pack('<i', len(main_targets)))
for main_target, idx in main_targets.iteritems():
LOG.verbose("{:,}\t{}\t# {:,}".format(len(main_target), main_target, idx))
f.write(pack('<i', len(main_target)))
f.write(pack('<{}s'.format(len(main_target)), main_target))
# haplotypes
LOG.verbose("{:,}\t# NUMBER OF HAPLOTYPES".format(len(haplotypes)))
f.write(pack('<i', len(haplotypes)))
for idx, hap in enumerate(haplotypes):
LOG.verbose("{:,}\t{}\t# {:,}".format(len(hap), hap, idx))
f.write(pack('<i', len(hap)))
f.write(pack('<{}s'.format(len(hap)), hap))
# equivalence classes
LOG.verbose("{:,}\t# NUMBER OF EQUIVALANCE CLASSES".format(len(ec)))
f.write(pack('<i', len(ec)))
for idx, k in enumerate(ec.keys()):
# ec[k] is the count
LOG.verbose("{:,}\t# {}\t{:,}".format(ec[k], k, idx))
f.write(pack('<i', ec[k]))
LOG.info("Determining mappings...")
# equivalence class mappings
counter = 0
for k, v in ec.iteritems():
arr_target_idx = k.split(",")
# get the main targets by name
temp_main_targets = set()
for idx in arr_target_idx:
temp_main_targets.add(target_idx_to_main_target[idx])
counter += len(temp_main_targets)
LOG.verbose("{:,}\t# NUMBER OF EQUIVALANCE CLASS MAPPINGS".format(counter))
f.write(pack('<i', counter))
for k, v in ec.iteritems():
arr_target_idx = k.split(",")
# get the main targets by name
temp_main_targets = set()
for idx in arr_target_idx:
temp_main_targets.add(target_idx_to_main_target[idx])
# loop through the haplotypes and targets to get the bits
for main_target in temp_main_targets:
# main_target is not an index, but a value like 'ENMUST..001'
bits = []
for hap in haplotypes:
read_transcript = '{}_{}'.format(main_target, hap) # now 'ENMUST..001_A'
read_transcript_idx = str(sam_file.gettid(read_transcript))
if read_transcript_idx in arr_target_idx:
bits.append(1)
else:
bits.append(0)
LOG.verbose("{}\t{}\t{}\t# {}\t{}".format(ec_idx[k], main_targets[main_target], list_to_int(bits), main_target, bits))
f.write(pack('<i', ec_idx[k]))
f.write(pack('<i', main_targets[main_target]))
f.write(pack('<i', list_to_int(bits)))
f.close()
except:
_show_error()
LOG.info("Done with converting BAM file!")
def emasify(binary_file_name, emase_file_name):
"""
:param binary_file_name:
:param emase_file_name:
:return:
"""
if not binary_file_name:
raise ValueError("empty file name, cannot load")
print "Binary File: {0}".format(binary_file_name)
f = open(binary_file_name, 'rb')
file_version = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
if file_version == 0:
print "Version: 0, Reads"
elif file_version == 1:
print "Version: 1, Equivalence Class"
else:
print "Unknown version, exiting"
# TARGETS
target_ids = []
targets = OrderedDict()
num_targets = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
print "Target Count: {0}".format(num_targets)
for i in xrange(0, num_targets):
str_len = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
target = np.fromfile(f, dtype=np.dtype('a' + str(str_len)), count=1)[0]
targets[target] = i
target_ids.append(target)
# HAPLOTYPES
haplotype_ids = []
haplotypes = OrderedDict()
num_haplotypes = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
print "Haplotype Count: {0}".format(num_haplotypes)
for i in xrange(0, num_haplotypes):
str_len = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
haplotype = np.fromfile(f, dtype=np.dtype('a' + str(str_len)),
count=1)[0]
haplotypes[haplotype] = i
haplotype_ids.append(haplotype)
if file_version == 0:
# READS
read_ids = []
reads = OrderedDict()
num_reads = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
print "Read Count: {0}".format(num_reads)
for i in xrange(0, num_reads):
str_len = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
read_id = np.fromfile(f,
dtype=np.dtype('a' + str(str_len)),
count=1)[0]
reads[read_id] = i
read_ids.append(read_id)
# ALIGNMENTS
num_alignments = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
print "Alignment Count: {0}".format(num_alignments)
alignments = np.fromfile(f,
dtype=np.dtype('i'),
count=num_alignments * 3)
print 'Creating APM...'
new_shape = (len(target_ids), len(haplotypes), len(read_ids))
aln_mat_kallisto = APM(shape=new_shape,
haplotype_names=haplotype_ids,
locus_names=target_ids,
read_names=read_ids)
print 'Parsing alignments...'
widgets = [Bar('>'), ' ', ETA(), ' ', Percentage()]
pbar = ProgressBar(widgets=widgets, maxval=num_alignments * 3).start()
counter = 0
for i in xrange(0, num_alignments * 3, 3):
rid = alignments[i]
lid = alignments[i + 1]
temp_bits = alignments[i + 2]
counter += 1
pbar.update(i)
if temp_bits == 0:
continue
bits = simple_from_one(temp_bits, num_haplotypes)
for hid, b in enumerate(bits):
if b:
aln_mat_kallisto.set_value(lid, hid, rid, 1)
pbar.finish()
print "Finalizing..."
aln_mat_kallisto.finalize()
aln_mat_kallisto.save(emase_file_name, title='KALLISTOALIGN')
print "DONE"
else:
# EQUIVALENCE CLASSES
num_ec = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
print "Equivalance Class Count: {0}".format(num_ec)
ec_ids = [x for x in xrange(0, num_ec)]
counts = np.fromfile(f, dtype=np.dtype('i'), count=num_ec)
# ALIGNMENTS
num_alignments = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
print "Alignment Count: {0}".format(num_alignments)
alignments = np.fromfile(f,
dtype=np.dtype('i'),
count=num_alignments * 3)
print 'Creating APM...'
new_shape = (len(target_ids), len(haplotypes), len(counts))
aln_mat_kallisto = APM(shape=new_shape,
haplotype_names=haplotype_ids,
locus_names=target_ids,
read_names=ec_ids)
aln_mat_kallisto.count = counts
print 'Parsing alignments...'
widgets = [Bar('>'), ' ', ETA(), ' ', Percentage()]
pbar = ProgressBar(widgets=widgets, maxval=num_alignments * 3).start()
counter = 0
for i in xrange(0, num_alignments * 3, 3):
rid = alignments[i]
lid = alignments[i + 1]
temp_bits = alignments[i + 2]
counter += 1
pbar.update(i)
if temp_bits == 0:
continue
bits = simple_from_one(temp_bits, num_haplotypes)
for hid, b in enumerate(bits):
if b:
aln_mat_kallisto.set_value(lid, hid, rid, 1)
pbar.finish()
print "Finalizing..."
aln_mat_kallisto.finalize()
aln_mat_kallisto.save(emase_file_name, title='KALLISTOALIGN')
print "DONE"
def bin2emase(binary_file_name, emase_file_name):
try:
if not binary_file_name:
raise ValueError("empty file name, cannot load")
LOG.info("Binary File: {0}".format(binary_file_name))
f = open(binary_file_name, 'rb')
file_version = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
if file_version == 0:
LOG.info("Version: 0, Reads, exiting")
sys.exit(-1)
elif file_version == 1:
LOG.info("Version: 1, Equivalence Class")
else:
LOG.info("Unknown version, exiting")
sys.exit(-1)
# TARGETS
target_ids = []
targets = OrderedDict()
num_targets = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
LOG.info("Target Count: {0:,}".format(num_targets))
for i in xrange(0, num_targets):
str_len = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
target = np.fromfile(f, dtype=np.dtype('a' + str(str_len)), count=1)[0]
targets[target] = i
target_ids.append(target)
LOG.verbose("{}\t{}".format(i, target))
# HAPLOTYPES
haplotype_ids = []
haplotypes = OrderedDict()
num_haplotypes = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
LOG.info("Haplotype Count: {0:,}".format(num_haplotypes))
for i in xrange(0, num_haplotypes):
str_len = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
haplotype = np.fromfile(f, dtype=np.dtype('a' + str(str_len)), count=1)[0]
haplotypes[haplotype] = i
haplotype_ids.append(haplotype)
LOG.verbose("{}\t{}".format(i, haplotype))
# EQUIVALENCE CLASSES
num_ec = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
LOG.info("Equivalance Class Count: {0:,}".format(num_ec))
ec_ids = [x for x in xrange(0, num_ec)]
counts = np.fromfile(f, dtype=np.dtype('i'), count=num_ec)
# ALIGNMENTS
num_alignments = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
LOG.info("Alignment Count: {0:,}".format(num_alignments))
new_shape = (num_targets, num_haplotypes, num_ec)
LOG.info('Creating APM...')
ec_ids = [x for x in xrange(0, num_ec)]
LOG.debug('Shape={}'.format(new_shape))
apm = APM(shape=new_shape, haplotype_names=haplotype_ids, locus_names=target_ids, read_names=ec_ids)
# counts -> the number of times this equivalence class has appeared
apm.count = counts
counter = 0
alignments = np.fromfile(f, dtype=np.dtype('i'), count=num_alignments*3)
for i in xrange(0, num_alignments*3, 3):
rid = alignments[i]
lid = alignments[i+1]
temp_bits = alignments[i+2]
counter += 1
if temp_bits == 0:
continue
try:
bits = int_to_list(temp_bits, num_haplotypes)
for i, bit in enumerate(bits):
if bit:
apm.set_value(rid, i, lid, 1)
except Exception, e:
_show_error()
raise e
LOG.info("Finalizing...")
apm.finalize()
apm.save(emase_file_name, title='bam2ec')
def emasify(binary_file_name, emase_file_name):
"""
:param binary_file_name:
:param emase_file_name:
:return:
"""
if not binary_file_name:
raise ValueError("empty file name, cannot load")
print "Binary File: {0}".format(binary_file_name)
f = open(binary_file_name, 'rb')
file_version = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
if file_version == 0:
print "Version: 0, Reads"
elif file_version == 1:
print "Version: 1, Equivalence Class"
else:
print "Unknown version, exiting"
# TARGETS
target_ids = []
targets = OrderedDict()
num_targets = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
print "Target Count: {0}".format(num_targets)
for i in xrange(0, num_targets):
str_len = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
target = np.fromfile(f, dtype=np.dtype('a' + str(str_len)), count=1)[0]
targets[target] = i
target_ids.append(target)
# HAPLOTYPES
haplotype_ids = []
haplotypes = OrderedDict()
num_haplotypes = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
print "Haplotype Count: {0}".format(num_haplotypes)
for i in xrange(0, num_haplotypes):
str_len = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
haplotype = np.fromfile(f, dtype=np.dtype('a' + str(str_len)), count=1)[0]
haplotypes[haplotype] = i
haplotype_ids.append(haplotype)
if file_version == 0:
# READS
read_ids = []
reads = OrderedDict()
num_reads = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
print "Read Count: {0}".format(num_reads)
for i in xrange(0, num_reads):
str_len = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
read_id = np.fromfile(f, dtype=np.dtype('a' + str(str_len)), count=1)[0]
reads[read_id] = i
read_ids.append(read_id)
# ALIGNMENTS
num_alignments = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
print "Alignment Count: {0}".format(num_alignments)
alignments = np.fromfile(f, dtype = np.dtype('i'), count=num_alignments*3)
print 'Creating APM...'
new_shape = (len(target_ids), len(haplotypes), len(read_ids))
aln_mat_kallisto = APM(shape=new_shape, haplotype_names=haplotype_ids, locus_names=target_ids, read_names=read_ids)
print 'Parsing alignments...'
widgets = [Bar('>'), ' ', ETA(), ' ', Percentage()]
pbar = ProgressBar(widgets=widgets, maxval=num_alignments*3).start()
counter = 0
for i in xrange(0, num_alignments*3, 3):
rid = alignments[i]
lid = alignments[i+1]
temp_bits = alignments[i+2]
counter += 1
pbar.update(i)
if temp_bits == 0:
continue
bits = simple_from_one(temp_bits, num_haplotypes)
for hid, b in enumerate(bits):
if b:
aln_mat_kallisto.set_value(lid, hid, rid, 1)
pbar.finish()
print "Finalizing..."
aln_mat_kallisto.finalize()
aln_mat_kallisto.save(emase_file_name, title='KALLISTOALIGN')
print "DONE"
else:
# EQUIVALENCE CLASSES
num_ec = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
print "Equivalance Class Count: {0}".format(num_ec)
ec_ids = [x for x in xrange(0, num_ec)]
counts = np.fromfile(f, dtype=np.dtype('i'), count=num_ec)
# ALIGNMENTS
num_alignments = np.fromfile(f, dtype=np.dtype('i'), count=1)[0]
print "Alignment Count: {0}".format(num_alignments)
alignments = np.fromfile(f, dtype = np.dtype('i'), count=num_alignments*3)
print 'Creating APM...'
new_shape = (len(target_ids), len(haplotypes), len(counts))
aln_mat_kallisto = APM(shape=new_shape, haplotype_names=haplotype_ids, locus_names=target_ids, read_names=ec_ids)
aln_mat_kallisto.count = counts
print 'Parsing alignments...'
widgets = [Bar('>'), ' ', ETA(), ' ', Percentage()]
pbar = ProgressBar(widgets=widgets, maxval=num_alignments*3).start()
counter = 0
for i in xrange(0, num_alignments*3, 3):
rid = alignments[i]
lid = alignments[i+1]
temp_bits = alignments[i+2]
counter += 1
pbar.update(i)
if temp_bits == 0:
continue
bits = simple_from_one(temp_bits, num_haplotypes)
for hid, b in enumerate(bits):
if b:
aln_mat_kallisto.set_value(lid, hid, rid, 1)
pbar.finish()
print "Finalizing..."
aln_mat_kallisto.finalize()
aln_mat_kallisto.save(emase_file_name, title='KALLISTOALIGN')
print "DONE"
