def run_signalp(fasta_files, output_file, tmp_dir):
def _consumer(worker_index, worker_fasta_file, output_queue):
# process with signalp
logging.debug('[Worker %d] Running SignalP' % (worker_index))
args = ['signalp', '-f', 'short', '-t', 'euk', worker_fasta_file]
tmp_output_file = os.path.join(tmp_dir, 'worker%d.signalp.txt' % (worker_index))
with open(tmp_output_file, 'w') as outfileh:
retcode = subprocess.call(args, stdout=outfileh)
if retcode != 0:
logging.error('[Worker %d] Error running SignalP' % (worker_index))
# put results into output queue
with open(tmp_output_file) as f:
for line in f:
if not line:
continue
line = line.strip()
if not line:
continue
if line.startswith('#'):
continue
output_queue.put(line)
output_queue.put(None)
logging.debug('[Worker %d] Finished' % (worker_index))
# create multiprocessing queue for passing data
output_queue = Queue(maxsize=len(fasta_files)*4)
# start consumer processes
procs = []
for i in xrange(len(fasta_files)):
p = Process(target=_consumer, args=(i, fasta_files[i], output_queue))
p.start()
procs.append(p)
# get results from consumers
num_alive = len(procs)
tmp_output_file = os.path.join(tmp_dir, 'signalp.merge.txt')
with open(tmp_output_file, 'w') as f:
while num_alive > 0:
result = output_queue.get()
if result is None:
num_alive -= 1
logging.debug("Main process detected worker finished, %d still alive" % (num_alive))
else:
print >>f, result
logging.debug("Joining all processes")
# wait for consumers to finish
for p in procs:
p.join()
# sort
logging.debug('Sorting signal peptide results')
def sort_signalp(line):
return line.split(None,1)[0]
sort_tmp_dir = os.path.join(tmp_dir, 'sort_tmp')
os.makedirs(sort_tmp_dir)
batch_sort(input=tmp_output_file,
output=output_file,
key=sort_signalp,
buffer_size=SORT_BUFFER_SIZE,
tempdirs=[sort_tmp_dir])
# cleanup temporary files
shutil.rmtree(sort_tmp_dir)
return 0
def sort_classification_results(input_file, output_file, tmp_dir):
# sort classification results
def sort_by_chrom_start(line):
fields = line.strip().split('\t', 2)
if fields[0] == "chrom":
return chr(0), 0
return fields[0], int(fields[1])
batch_sort(input=input_file,
output=output_file,
key=sort_by_chrom_start,
buffer_size=(1 << 21),
tempdirs=[tmp_dir])
def sort_classification_results(input_file, output_file, tmp_dir):
# sort classification results
def sort_by_chrom_start(line):
fields = line.strip().split('\t', 2)
if fields[0] == "chrom":
return chr(0), 0
return fields[0], int(fields[1])
batch_sort(input=input_file,
output=output_file,
key=sort_by_chrom_start,
buffer_size=(1 << 21),
tempdirs=[tmp_dir])
def orf_analysis(gtf_file, genome_fasta_file, pfam_dir,
output_dir, min_orf_length, first_orf_only,
num_processes):
#
# extract transcript DNA sequences, translate to protein, and
# search for ORFs
#
logging.debug('Finding ORFs in transcript sequences')
# output files
tmp_dir = os.path.join(output_dir, 'tmp')
if not os.path.exists(tmp_dir):
logging.info("Creating tmp directory '%s'" % (tmp_dir))
os.makedirs(tmp_dir)
orf_bed_file = os.path.join(output_dir, 'transcript_orfs.bed')
unique_orf_file = os.path.join(output_dir, 'unique_orfs.txt')
unique_orf_bed_file = os.path.join(output_dir, 'unique_orfs.bed')
orf_file = os.path.join(tmp_dir, 'transcript_orfs.no_ids.txt')
sorted_orf_file = os.path.join(tmp_dir, 'transcript_orfs.no_ids.sortbyorf.txt')
sorted_orf_id_file = os.path.join(tmp_dir, 'transcript_orfs.sortbyorf.txt')
signalp_file = os.path.join(output_dir, 'signalp.txt')
pfam_file = os.path.join(output_dir, 'pfam.txt')
merged_orf_id_file = os.path.join(output_dir, 'transcript_orfs.sortbyorf.merged.txt')
sorted_merged_orf_id_file = os.path.join(output_dir, 'transcript_orfs.sortbytranscript.merged.txt')
# open output files
orf_fileh = open(orf_file, 'w')
orf_bed_fileh = open(orf_bed_file, 'w')
# open genome fasta file
ref_fa = pysam.Fastafile(genome_fasta_file)
num_finished = 1
for locus_transcripts in parse_gtf(open(gtf_file)):
for t in locus_transcripts:
if first_orf_only:
# get first ORF
orf = get_first_transcript_orf(t, ref_fa)
if len(orf.seq) >= min_orf_length:
print >>orf_fileh, '\t'.join(orf.to_table())
print >>orf_bed_fileh, '\t'.join(orf.to_bed())
else:
# get all ORFs
for orf in get_all_transcript_orfs(t, ref_fa, min_orf_length):
print >>orf_fileh, '\t'.join(orf.to_table())
print >>orf_bed_fileh, '\t'.join(orf.to_bed())
if (num_finished % 10000) == 0:
logging.debug('Processed %d transcripts' % (num_finished))
num_finished += 1
# cleanup
orf_fileh.close()
orf_bed_fileh.close()
#
# sort ORF table by ORF amino acid sequence to group identical ORFs
# together
#
logging.debug('Sorting ORFs by amino acid sequence')
def sort_by_seq(line):
'''comparison function for batch_sort'''
fields = line.strip().split('\t')
return fields[ORFInfo.SEQ_COL_NUM]
sort_tmp_dir = os.path.join(tmp_dir, 'sort_tmp')
os.makedirs(sort_tmp_dir)
batch_sort(input=orf_file,
output=sorted_orf_file,
key=sort_by_seq,
buffer_size=SORT_BUFFER_SIZE,
tempdirs=[sort_tmp_dir])
shutil.rmtree(sort_tmp_dir)
#
# assign each ORF a unique id and write to FASTA file
#
logging.debug('Determining unique ORFs')
orf_fasta_prefix = os.path.join(tmp_dir, 'orf')
orf_fasta_files = []
orf_fasta_sizes = []
for i in xrange(num_processes):
orf_fasta_files.append(open('%s%d.fasta' % (orf_fasta_prefix, i), 'w'))
orf_fasta_sizes.append(0)
orf_file_index = 0
outfileh = open(sorted_orf_id_file, 'w')
unique_orf_fileh = open(unique_orf_file, 'w')
print >>unique_orf_fileh, '\t'.join(['orf_id', 'orf_length', 'total_occurrences', 'unique_genomic_occurrences'])
unique_orf_bed_fileh = open(unique_orf_bed_file, 'w')
with open(sorted_orf_file) as infileh:
for orfs in group_unique_orfs(infileh):
# write to master transcript/ORF table
for orf in orfs:
print >>outfileh, '\t'.join(orf.to_table())
# write ORF to fasta file
lines = to_fasta(orfs[0].orf_id, orfs[0].seq.strip('*'))
print >>orf_fasta_files[orf_file_index], lines
orf_fasta_sizes[orf_file_index] += 1
# advance to next fasta file
orf_file_index = (orf_file_index + 1) % (num_processes)
# group by genomic position and write ORFs to BED file
unique_genome_orfs = {}
for orf in orfs:
k = (orf.chrom, orf.strand, tuple(orf.exons))
if k in unique_genome_orfs:
continue
unique_genome_orfs[k] = orf
for orf in unique_genome_orfs.itervalues():
print >>unique_orf_bed_fileh, '\t'.join(orf.to_bed(orf.orf_id))
# write unique ORF to tab-delimited text file
fields = [orfs[0].orf_id, len(orfs[0].seq), len(orfs), len(unique_genome_orfs)]
print >>unique_orf_fileh, '\t'.join(map(str,fields))
# cleanup
unique_orf_bed_fileh.close()
outfileh.close()
# get fasta files with lines written
orf_fasta_file_names = []
for i in xrange(len(orf_fasta_files)):
orf_fasta_files[i].close()
if orf_fasta_sizes[i] > 0:
orf_fasta_file_names.append(orf_fasta_files[i].name)
#
# search FASTA file against signalp
#
logging.debug('Searching for signal peptides')
retcode = run_signalp(orf_fasta_file_names, signalp_file, tmp_dir)
if retcode != 0:
logging.error('Error searching for signal peptides')
return 1
#
# search FASTA file against Pfam
#
logging.error('Scanning for Pfam domains')
retcode = run_pfam(orf_fasta_file_names, pfam_dir, pfam_file, tmp_dir)
if retcode != 0:
logging.error('Error running pfam_scan.pl')
#
# merge results from Pfam and signalp
#
logging.debug('Merging SignalP and Pfam results')
merge_results(sorted_orf_id_file, signalp_file, pfam_file, merged_orf_id_file)
#
# sort by transcript id
#
logging.debug('Sorting ORFs by transcript ID')
def sort_by_transcript_id(line):
return line.split('\t', 1)[0]
sort_tmp_dir = os.path.join(tmp_dir, 'sort_tmp')
os.makedirs(sort_tmp_dir)
batch_sort(input=merged_orf_id_file,
output=sorted_merged_orf_id_file,
key=sort_by_transcript_id,
buffer_size=SORT_BUFFER_SIZE,
tempdirs=[sort_tmp_dir])
shutil.rmtree(sort_tmp_dir)
# cleanup
ref_fa.close()
return 0
def orf_analysis(gtf_file, genome_fasta_file, pfam_dir, output_dir,
min_orf_length, first_orf_only, num_processes):
#
# extract transcript DNA sequences, translate to protein, and
# search for ORFs
#
logging.debug('Finding ORFs in transcript sequences')
# output files
tmp_dir = os.path.join(output_dir, 'tmp')
if not os.path.exists(tmp_dir):
logging.info("Creating tmp directory '%s'" % (tmp_dir))
os.makedirs(tmp_dir)
orf_bed_file = os.path.join(output_dir, 'transcript_orfs.bed')
unique_orf_file = os.path.join(output_dir, 'unique_orfs.txt')
unique_orf_bed_file = os.path.join(output_dir, 'unique_orfs.bed')
orf_file = os.path.join(tmp_dir, 'transcript_orfs.no_ids.txt')
sorted_orf_file = os.path.join(tmp_dir,
'transcript_orfs.no_ids.sortbyorf.txt')
sorted_orf_id_file = os.path.join(tmp_dir, 'transcript_orfs.sortbyorf.txt')
signalp_file = os.path.join(output_dir, 'signalp.txt')
pfam_file = os.path.join(output_dir, 'pfam.txt')
merged_orf_id_file = os.path.join(output_dir,
'transcript_orfs.sortbyorf.merged.txt')
sorted_merged_orf_id_file = os.path.join(
output_dir, 'transcript_orfs.sortbytranscript.merged.txt')
# open output files
orf_fileh = open(orf_file, 'w')
orf_bed_fileh = open(orf_bed_file, 'w')
# open genome fasta file
ref_fa = pysam.Fastafile(genome_fasta_file)
num_finished = 1
for locus_transcripts in parse_gtf(open(gtf_file)):
for t in locus_transcripts:
if first_orf_only:
# get first ORF
orf = get_first_transcript_orf(t, ref_fa)
if len(orf.seq) >= min_orf_length:
print >> orf_fileh, '\t'.join(orf.to_table())
print >> orf_bed_fileh, '\t'.join(orf.to_bed())
else:
# get all ORFs
for orf in get_all_transcript_orfs(t, ref_fa, min_orf_length):
print >> orf_fileh, '\t'.join(orf.to_table())
print >> orf_bed_fileh, '\t'.join(orf.to_bed())
if (num_finished % 10000) == 0:
logging.debug('Processed %d transcripts' % (num_finished))
num_finished += 1
# cleanup
orf_fileh.close()
orf_bed_fileh.close()
#
# sort ORF table by ORF amino acid sequence to group identical ORFs
# together
#
logging.debug('Sorting ORFs by amino acid sequence')
def sort_by_seq(line):
'''comparison function for batch_sort'''
fields = line.strip().split('\t')
return fields[ORFInfo.SEQ_COL_NUM]
sort_tmp_dir = os.path.join(tmp_dir, 'sort_tmp')
os.makedirs(sort_tmp_dir)
batch_sort(input=orf_file,
output=sorted_orf_file,
key=sort_by_seq,
buffer_size=SORT_BUFFER_SIZE,
tempdirs=[sort_tmp_dir])
shutil.rmtree(sort_tmp_dir)
#
# assign each ORF a unique id and write to FASTA file
#
logging.debug('Determining unique ORFs')
orf_fasta_prefix = os.path.join(tmp_dir, 'orf')
orf_fasta_files = []
orf_fasta_sizes = []
for i in xrange(num_processes):
orf_fasta_files.append(open('%s%d.fasta' % (orf_fasta_prefix, i), 'w'))
orf_fasta_sizes.append(0)
orf_file_index = 0
outfileh = open(sorted_orf_id_file, 'w')
unique_orf_fileh = open(unique_orf_file, 'w')
print >> unique_orf_fileh, '\t'.join([
'orf_id', 'orf_length', 'total_occurrences',
'unique_genomic_occurrences'
])
unique_orf_bed_fileh = open(unique_orf_bed_file, 'w')
with open(sorted_orf_file) as infileh:
for orfs in group_unique_orfs(infileh):
# write to master transcript/ORF table
for orf in orfs:
print >> outfileh, '\t'.join(orf.to_table())
# write ORF to fasta file
lines = to_fasta(orfs[0].orf_id, orfs[0].seq.strip('*'))
print >> orf_fasta_files[orf_file_index], lines
orf_fasta_sizes[orf_file_index] += 1
# advance to next fasta file
orf_file_index = (orf_file_index + 1) % (num_processes)
# group by genomic position and write ORFs to BED file
unique_genome_orfs = {}
for orf in orfs:
k = (orf.chrom, orf.strand, tuple(orf.exons))
if k in unique_genome_orfs:
continue
unique_genome_orfs[k] = orf
for orf in unique_genome_orfs.itervalues():
print >> unique_orf_bed_fileh, '\t'.join(orf.to_bed(
orf.orf_id))
# write unique ORF to tab-delimited text file
fields = [
orfs[0].orf_id,
len(orfs[0].seq),
len(orfs),
len(unique_genome_orfs)
]
print >> unique_orf_fileh, '\t'.join(map(str, fields))
# cleanup
unique_orf_bed_fileh.close()
outfileh.close()
# get fasta files with lines written
orf_fasta_file_names = []
for i in xrange(len(orf_fasta_files)):
orf_fasta_files[i].close()
if orf_fasta_sizes[i] > 0:
orf_fasta_file_names.append(orf_fasta_files[i].name)
#
# search FASTA file against signalp
#
logging.debug('Searching for signal peptides')
retcode = run_signalp(orf_fasta_file_names, signalp_file, tmp_dir)
if retcode != 0:
logging.error('Error searching for signal peptides')
return 1
#
# search FASTA file against Pfam
#
logging.error('Scanning for Pfam domains')
retcode = run_pfam(orf_fasta_file_names, pfam_dir, pfam_file, tmp_dir)
if retcode != 0:
logging.error('Error running pfam_scan.pl')
#
# merge results from Pfam and signalp
#
logging.debug('Merging SignalP and Pfam results')
merge_results(sorted_orf_id_file, signalp_file, pfam_file,
merged_orf_id_file)
#
# sort by transcript id
#
logging.debug('Sorting ORFs by transcript ID')
def sort_by_transcript_id(line):
return line.split('\t', 1)[0]
sort_tmp_dir = os.path.join(tmp_dir, 'sort_tmp')
os.makedirs(sort_tmp_dir)
batch_sort(input=merged_orf_id_file,
output=sorted_merged_orf_id_file,
key=sort_by_transcript_id,
buffer_size=SORT_BUFFER_SIZE,
tempdirs=[sort_tmp_dir])
shutil.rmtree(sort_tmp_dir)
# cleanup
ref_fa.close()
return 0
def run_pfam(fasta_files, pfam_dir, output_file, tmp_dir):
def _consumer(worker_index, worker_fasta_file, output_queue):
# process with pfam
logging.debug('[Worker %d] Running pfam_scan.pl' % (worker_index))
tmp_output_file = os.path.join(tmp_dir,
'worker%d.pfam.txt' % (worker_index))
args = [
'pfam_scan.pl', '-cpu', '1', '-pfamB', '-fasta', worker_fasta_file,
'-dir', pfam_dir, '-outfile', tmp_output_file
]
retcode = subprocess.call(args)
if retcode != 0:
logging.error('[Worker %d] Error running pfam_scan.pl' %
(worker_index))
# put results into output queue
with open(tmp_output_file) as f:
for line in f:
if not line:
continue
line = line.strip()
if not line:
continue
if line.startswith('#'):
continue
output_queue.put(line)
output_queue.put(None)
logging.debug('[Worker %d] Finished' % (worker_index))
# create multiprocessing queues for passing data
output_queue = Queue(maxsize=len(fasta_files) * 4)
# start consumer processes
procs = []
for i in xrange(len(fasta_files)):
p = Process(target=_consumer, args=(i, fasta_files[i], output_queue))
p.start()
procs.append(p)
# get results from consumers
num_alive = len(procs)
tmp_output_file = os.path.join(tmp_dir, 'pfam.merge.txt')
with open(tmp_output_file, 'w') as f:
while num_alive > 0:
result = output_queue.get()
if result is None:
num_alive -= 1
logging.debug(
"Main process detected worker finished, %d still alive" %
(num_alive))
else:
print >> f, result
logging.debug("Joining all processes")
# wait for consumers to finish
for p in procs:
p.join()
# sort
logging.debug('Sorting pfam results')
def sort_pfam(line):
return line.split(None, 1)[0]
sort_tmp_dir = os.path.join(tmp_dir, 'sort_tmp')
os.makedirs(sort_tmp_dir)
batch_sort(input=tmp_output_file,
output=output_file,
key=sort_pfam,
buffer_size=SORT_BUFFER_SIZE,
tempdirs=[sort_tmp_dir])
# cleanup temporary files
shutil.rmtree(sort_tmp_dir)
return 0
