Mercurial > repos > xuebing > sharplabtool
diff tools/ilmn_pacbio/quake.py @ 0:9071e359b9a3
Uploaded
author | xuebing |
---|---|
date | Fri, 09 Mar 2012 19:37:19 -0500 |
parents | |
children |
line wrap: on
line diff
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/tools/ilmn_pacbio/quake.py Fri Mar 09 19:37:19 2012 -0500 @@ -0,0 +1,136 @@ +#!/usr/bin/env python +from optparse import OptionParser, SUPPRESS_HELP +import os, random, sys +import cov_model + +############################################################ +# quake.py +# +# Launch pipeline to correct errors in Illumina sequencing +# reads. +############################################################ + +#r_dir = '/nfshomes/dakelley/research/error_correction/bin' +quake_dir = os.path.abspath(os.path.dirname(sys.argv[0])) + +############################################################ +# main +############################################################ +def main(): + usage = 'usage: %prog [options]' + parser = OptionParser(usage) + parser.add_option('-r', dest='readsf', help='Fastq file of reads') + parser.add_option('-f', dest='reads_listf', help='File containing fastq file names, one per line or two per line for paired end reads.') + parser.add_option('-k', dest='k', type='int', help='Size of k-mers to correct') + parser.add_option('-p', dest='proc', type='int', default=4, help='Number of processes [default: %default]') + parser.add_option('-q', dest='quality_scale', type='int', default=-1, help='Quality value ascii scale, generally 64 or 33. If not specified, it will guess.') + parser.add_option('--no_count', dest='no_count', action='store_true', default=False, help='Kmers are already counted and in expected file [reads file].qcts or [reads file].cts [default: %default]') + parser.add_option('--no_cut', dest='no_cut', action='store_true', default=False, help='Coverage model is optimized and cutoff was printed to expected file cutoff.txt [default: %default]') + parser.add_option('--int', dest='counted_kmers', action='store_true', default=False, help='Kmers were counted as integers w/o the use of quality values [default: %default]') + parser.add_option('--ratio', dest='ratio', type='int', default=200, help='Likelihood ratio to set trusted/untrusted cutoff. Generally set between 10-1000 with lower numbers suggesting a lower threshold. [default: %default]') + # help='Model kmer coverage as a function of GC content of kmers [default: %default]' + parser.add_option('--gc', dest='model_gc', action='store_true', default=False, help=SUPPRESS_HELP) + parser.add_option('--headers', action='store_true', default=False, help='Output original read headers (i.e. pass --headers to correct)' ) + (options, args) = parser.parse_args() + + if not options.readsf and not options.reads_listf: + parser.error('Must provide fastq file of reads with -r or file with list of fastq files of reads with -f') + if not options.k: + parser.error('Must provide k-mer size with -k') + if options.quality_scale == -1: + options.quality_scale = guess_quality_scale(options.readsf, options.reads_listf) + + if options.counted_kmers: + cts_suf = 'cts' + else: + cts_suf = 'qcts' + if options.readsf: + ctsf = '%s.%s' % (os.path.splitext( os.path.split(options.readsf)[1] )[0], cts_suf) + reads_str = '-r %s' % options.readsf + else: + ctsf = '%s.%s' % (os.path.split(options.reads_listf)[1], cts_suf) + reads_str = '-f %s' % options.reads_listf + + if not options.no_count and not options.no_cut: + count_kmers(options.readsf, options.reads_listf, options.k, ctsf, options.quality_scale) + + if not options.no_cut: + # model coverage + if options.counted_kmers: + cov_model.model_cutoff(ctsf, options.ratio) + else: + if options.model_gc: + cov_model.model_q_gc_cutoffs(ctsf, 10000, options.ratio) + else: + cov_model.model_q_cutoff(ctsf, 25000, options.ratio) + + + if options.model_gc: + # run correct C++ code + os.system('%s/correct %s -k %d -m %s -a cutoffs.gc.txt -p %d -q %d' % (quake_dir,reads_str, options.k, ctsf, options.proc, options.quality_scale)) + + else: + cutoff = open('cutoff.txt').readline().rstrip() + + # run correct C++ code + headers = '--headers' if options.headers else '' + os.system('%s/correct %s %s -k %d -m %s -c %s -p %d -q %d' % (quake_dir,headers, reads_str, options.k, ctsf, cutoff, options.proc, options.quality_scale)) + + +################################################################################ +# guess_quality_scale +# Guess at ascii scale of quality values by examining +# a bunch of reads and looking for quality values < 64, +# in which case we set it to 33. +################################################################################ +def guess_quality_scale(readsf, reads_listf): + reads_to_check = 1000 + if not readsf: + readsf = open(reads_listf).readline().split()[0] + + fqf = open(readsf) + reads_checked = 0 + header = fqf.readline() + while header and reads_checked < reads_to_check: + seq = fqf.readline() + mid = fqf.readline() + qual = fqf.readline().rstrip() + reads_checked += 1 + for q in qual: + if ord(q) < 64: + print 'Guessing quality values are on ascii 33 scale' + return 33 + header = fqf.readline() + + print 'Guessing quality values are on ascii 64 scale' + return 64 + + + +############################################################ +# count_kmers +# +# Count kmers in the reads file using AMOS count-kmers or +# count-qmers +############################################################ +def count_kmers(readsf, reads_listf, k, ctsf, quality_scale): + # find files + fq_files = [] + if readsf: + fq_files.append(readsf) + else: + for line in open(reads_listf): + for fqf in line.split(): + fq_files.append(fqf) + + if ctsf[-4:] == 'qcts': + os.system('cat %s | %s/count-qmers -k %d -q %d > %s' % (' '.join(fq_files), quake_dir, k, quality_scale, ctsf)) + else: + os.system('cat %s | %s/count-kmers -k %d > %s' % (' '.join(fq_files), quake_dir, k, ctsf)) + + +############################################################ +# __main__ +############################################################ +if __name__ == '__main__': + main()