Mercurial > repos > curtish > bwa_tools_mini
view bwa_wrapper.py @ 0:11eda1d95d84 draft default tip
clone upstream
author | curtish |
---|---|
date | Mon, 26 Aug 2013 17:06:46 -0400 |
parents | |
children |
line wrap: on
line source
#!/usr/bin/env python """ Runs BWA on single-end or paired-end data. Produces a SAM file containing the mappings. Works with BWA version 0.5.9. usage: bwa_wrapper.py [options] See below for options """ import optparse, os, shutil, subprocess, sys, tempfile def stop_err( msg ): sys.stderr.write( '%s\n' % msg ) sys.exit() def check_is_double_encoded( fastq ): # check that first read is bases, not one base followed by numbers bases = [ 'A', 'C', 'G', 'T', 'a', 'c', 'g', 't', 'N' ] nums = [ '0', '1', '2', '3' ] for line in file( fastq, 'rb'): if not line.strip() or line.startswith( '@' ): continue if len( [ b for b in line.strip() if b in nums ] ) > 0: return False elif line.strip()[0] in bases and len( [ b for b in line.strip() if b in bases ] ) == len( line.strip() ): return True else: raise Exception, 'First line in first read does not appear to be a valid FASTQ read in either base-space or color-space' raise Exception, 'There is no non-comment and non-blank line in your FASTQ file' def __main__(): #Parse Command Line parser = optparse.OptionParser() parser.add_option( '-t', '--threads', dest='threads', help='The number of threads to use' ) parser.add_option( '-c', '--color-space', dest='color_space', action='store_true', help='If the input files are SOLiD format' ) parser.add_option( '-r', '--ref', dest='ref', help='The reference genome to use or index' ) parser.add_option( '-f', '--input1', dest='fastq', help='The (forward) fastq file to use for the mapping' ) parser.add_option( '-F', '--input2', dest='rfastq', help='The reverse fastq file to use for mapping if paired-end data' ) parser.add_option( '-u', '--output', dest='output', help='The file to save the output (SAM format)' ) parser.add_option( '-g', '--genAlignType', dest='genAlignType', help='The type of pairing (single or paired)' ) parser.add_option( '-p', '--params', dest='params', help='Parameter setting to use (pre_set or full)' ) parser.add_option( '-s', '--fileSource', dest='fileSource', help='Whether to use a previously indexed reference sequence or one form history (indexed or history)' ) parser.add_option( '-n', '--maxEditDist', dest='maxEditDist', help='Maximum edit distance if integer' ) parser.add_option( '-m', '--fracMissingAligns', dest='fracMissingAligns', help='Fraction of missing alignments given 2% uniform base error rate if fraction' ) parser.add_option( '-o', '--maxGapOpens', dest='maxGapOpens', help='Maximum number of gap opens' ) parser.add_option( '-e', '--maxGapExtens', dest='maxGapExtens', help='Maximum number of gap extensions' ) parser.add_option( '-d', '--disallowLongDel', dest='disallowLongDel', help='Disallow a long deletion within specified bps' ) parser.add_option( '-i', '--disallowIndel', dest='disallowIndel', help='Disallow indel within specified bps' ) parser.add_option( '-l', '--seed', dest='seed', help='Take the first specified subsequences' ) parser.add_option( '-k', '--maxEditDistSeed', dest='maxEditDistSeed', help='Maximum edit distance to the seed' ) parser.add_option( '-M', '--mismatchPenalty', dest='mismatchPenalty', help='Mismatch penalty' ) parser.add_option( '-O', '--gapOpenPenalty', dest='gapOpenPenalty', help='Gap open penalty' ) parser.add_option( '-E', '--gapExtensPenalty', dest='gapExtensPenalty', help='Gap extension penalty' ) parser.add_option( '-R', '--suboptAlign', dest='suboptAlign', default=None, help='Proceed with suboptimal alignments even if the top hit is a repeat' ) parser.add_option( '-N', '--noIterSearch', dest='noIterSearch', help='Disable iterative search' ) parser.add_option( '-T', '--outputTopN', dest='outputTopN', help='Maximum number of alignments to output in the XA tag for reads paired properly' ) parser.add_option( '', '--outputTopNDisc', dest='outputTopNDisc', help='Maximum number of alignments to output in the XA tag for disconcordant read pairs (excluding singletons)' ) parser.add_option( '-S', '--maxInsertSize', dest='maxInsertSize', help='Maximum insert size for a read pair to be considered mapped good' ) parser.add_option( '-P', '--maxOccurPairing', dest='maxOccurPairing', help='Maximum occurrences of a read for pairings' ) parser.add_option( '', '--rgid', dest='rgid', help='Read group identifier' ) parser.add_option( '', '--rgcn', dest='rgcn', help='Sequencing center that produced the read' ) parser.add_option( '', '--rgds', dest='rgds', help='Description' ) parser.add_option( '', '--rgdt', dest='rgdt', help='Date that run was produced (ISO8601 format date or date/time, like YYYY-MM-DD)' ) parser.add_option( '', '--rgfo', dest='rgfo', help='Flow order' ) parser.add_option( '', '--rgks', dest='rgks', help='The array of nucleotide bases that correspond to the key sequence of each read' ) parser.add_option( '', '--rglb', dest='rglb', help='Library name' ) parser.add_option( '', '--rgpg', dest='rgpg', help='Programs used for processing the read group' ) parser.add_option( '', '--rgpi', dest='rgpi', help='Predicted median insert size' ) parser.add_option( '', '--rgpl', dest='rgpl', choices=[ 'CAPILLARY', 'LS454', 'ILLUMINA', 'SOLID', 'HELICOS', 'IONTORRENT' and 'PACBIO' ], help='Platform/technology used to produce the reads' ) parser.add_option( '', '--rgpu', dest='rgpu', help='Platform unit (e.g. flowcell-barcode.lane for Illumina or slide for SOLiD)' ) parser.add_option( '', '--rgsm', dest='rgsm', help='Sample' ) parser.add_option( '-D', '--dbkey', dest='dbkey', help='Dbkey for reference genome' ) parser.add_option( '-X', '--do_not_build_index', dest='do_not_build_index', action='store_true', help="Don't build index" ) parser.add_option( '-H', '--suppressHeader', dest='suppressHeader', help='Suppress header' ) parser.add_option( '-I', '--illumina1.3', dest='illumina13qual', help='Input FASTQ files have Illuina 1.3 quality scores' ) (options, args) = parser.parse_args() # output version # of tool try: tmp = tempfile.NamedTemporaryFile().name tmp_stdout = open( tmp, 'wb' ) proc = subprocess.Popen( args='bwa 2>&1', shell=True, stdout=tmp_stdout ) tmp_stdout.close() returncode = proc.wait() stdout = None for line in open( tmp_stdout.name, 'rb' ): if line.lower().find( 'version' ) >= 0: stdout = line.strip() break if stdout: sys.stdout.write( 'BWA %s\n' % stdout ) else: raise Exception except: sys.stdout.write( 'Could not determine BWA version\n' ) # check for color space fastq that's not double-encoded and exit if appropriate if options.color_space: if not check_is_double_encoded( options.fastq ): stop_err( 'Your file must be double-encoded (it must be converted from "numbers" to "bases"). See the help section for details' ) if options.genAlignType == 'paired': if not check_is_double_encoded( options.rfastq ): stop_err( 'Your reverse reads file must also be double-encoded (it must be converted from "numbers" to "bases"). See the help section for details' ) fastq = options.fastq if options.rfastq: rfastq = options.rfastq # set color space variable if options.color_space: color_space = '-c' else: color_space = '' # make temp directory for placement of indices tmp_index_dir = tempfile.mkdtemp() tmp_dir = tempfile.mkdtemp() # index if necessary if options.fileSource == 'history' and not options.do_not_build_index: ref_file = tempfile.NamedTemporaryFile( dir=tmp_index_dir ) ref_file_name = ref_file.name ref_file.close() os.symlink( options.ref, ref_file_name ) # determine which indexing algorithm to use, based on size try: size = os.stat( options.ref ).st_size if size <= 2**30: indexingAlg = 'is' else: indexingAlg = 'bwtsw' except: indexingAlg = 'is' indexing_cmds = '%s -a %s' % ( color_space, indexingAlg ) cmd1 = 'bwa index %s %s' % ( indexing_cmds, ref_file_name ) try: tmp = tempfile.NamedTemporaryFile( dir=tmp_index_dir ).name tmp_stderr = open( tmp, 'wb' ) proc = subprocess.Popen( args=cmd1, shell=True, cwd=tmp_index_dir, stderr=tmp_stderr.fileno() ) returncode = proc.wait() tmp_stderr.close() # get stderr, allowing for case where it's very large tmp_stderr = open( tmp, 'rb' ) stderr = '' buffsize = 1048576 try: while True: stderr += tmp_stderr.read( buffsize ) if not stderr or len( stderr ) % buffsize != 0: break except OverflowError: pass tmp_stderr.close() if returncode != 0: raise Exception, stderr except Exception, e: # clean up temp dirs if os.path.exists( tmp_index_dir ): shutil.rmtree( tmp_index_dir ) if os.path.exists( tmp_dir ): shutil.rmtree( tmp_dir ) stop_err( 'Error indexing reference sequence. ' + str( e ) ) else: ref_file_name = options.ref if options.illumina13qual: illumina_quals = "-I" else: illumina_quals = "" # set up aligning and generate aligning command options if options.params == 'pre_set': aligning_cmds = '-t %s %s %s' % ( options.threads, color_space, illumina_quals ) gen_alignment_cmds = '' else: if options.maxEditDist != '0': editDist = options.maxEditDist else: editDist = options.fracMissingAligns if options.seed != '-1': seed = '-l %s' % options.seed else: seed = '' if options.suboptAlign: suboptAlign = '-R "%s"' % ( options.suboptAlign ) else: suboptAlign = '' if options.noIterSearch == 'true': noIterSearch = '-N' else: noIterSearch = '' aligning_cmds = '-n %s -o %s -e %s -d %s -i %s %s -k %s -t %s -M %s -O %s -E %s %s %s %s %s' % \ ( editDist, options.maxGapOpens, options.maxGapExtens, options.disallowLongDel, options.disallowIndel, seed, options.maxEditDistSeed, options.threads, options.mismatchPenalty, options.gapOpenPenalty, options.gapExtensPenalty, suboptAlign, noIterSearch, color_space, illumina_quals ) if options.genAlignType == 'paired': gen_alignment_cmds = '-a %s -o %s' % ( options.maxInsertSize, options.maxOccurPairing ) if options.outputTopNDisc: gen_alignment_cmds += ' -N %s' % options.outputTopNDisc else: gen_alignment_cmds = '' if options.rgid: if not options.rglb or not options.rgpl or not options.rgsm: stop_err( 'If you want to specify read groups, you must include the ID, LB, PL, and SM tags.' ) readGroup = '@RG\tID:%s\tLB:%s\tPL:%s\tSM:%s' % ( options.rgid, options.rglb, options.rgpl, options.rgsm ) if options.rgcn: readGroup += '\tCN:%s' % options.rgcn if options.rgds: readGroup += '\tDS:%s' % options.rgds if options.rgdt: readGroup += '\tDT:%s' % options.rgdt if options.rgfo: readGroup += '\tFO:%s' % options.rgfo if options.rgks: readGroup += '\tKS:%s' % options.rgks if options.rgpg: readGroup += '\tPG:%s' % options.rgpg if options.rgpi: readGroup += '\tPI:%s' % options.rgpi if options.rgpu: readGroup += '\tPU:%s' % options.rgpu gen_alignment_cmds += ' -r "%s"' % readGroup if options.outputTopN: gen_alignment_cmds += ' -n %s' % options.outputTopN # set up output files tmp_align_out = tempfile.NamedTemporaryFile( dir=tmp_dir ) tmp_align_out_name = tmp_align_out.name tmp_align_out.close() tmp_align_out2 = tempfile.NamedTemporaryFile( dir=tmp_dir ) tmp_align_out2_name = tmp_align_out2.name tmp_align_out2.close() # prepare actual aligning and generate aligning commands cmd2 = 'bwa aln %s %s %s > %s' % ( aligning_cmds, ref_file_name, fastq, tmp_align_out_name ) cmd2b = '' if options.genAlignType == 'paired': cmd2b = 'bwa aln %s %s %s > %s' % ( aligning_cmds, ref_file_name, rfastq, tmp_align_out2_name ) cmd3 = 'bwa sampe %s %s %s %s %s %s >> %s' % ( gen_alignment_cmds, ref_file_name, tmp_align_out_name, tmp_align_out2_name, fastq, rfastq, options.output ) else: cmd3 = 'bwa samse %s %s %s %s >> %s' % ( gen_alignment_cmds, ref_file_name, tmp_align_out_name, fastq, options.output ) # perform alignments buffsize = 1048576 try: # need to nest try-except in try-finally to handle 2.4 try: # align try: tmp = tempfile.NamedTemporaryFile( dir=tmp_dir ).name tmp_stderr = open( tmp, 'wb' ) proc = subprocess.Popen( args=cmd2, shell=True, cwd=tmp_dir, stderr=tmp_stderr.fileno() ) returncode = proc.wait() tmp_stderr.close() # get stderr, allowing for case where it's very large tmp_stderr = open( tmp, 'rb' ) stderr = '' try: while True: stderr += tmp_stderr.read( buffsize ) if not stderr or len( stderr ) % buffsize != 0: break except OverflowError: pass tmp_stderr.close() if returncode != 0: raise Exception, stderr except Exception, e: raise Exception, 'Error aligning sequence. ' + str( e ) # and again if paired data try: if cmd2b: tmp = tempfile.NamedTemporaryFile( dir=tmp_dir ).name tmp_stderr = open( tmp, 'wb' ) proc = subprocess.Popen( args=cmd2b, shell=True, cwd=tmp_dir, stderr=tmp_stderr.fileno() ) returncode = proc.wait() tmp_stderr.close() # get stderr, allowing for case where it's very large tmp_stderr = open( tmp, 'rb' ) stderr = '' try: while True: stderr += tmp_stderr.read( buffsize ) if not stderr or len( stderr ) % buffsize != 0: break except OverflowError: pass tmp_stderr.close() if returncode != 0: raise Exception, stderr except Exception, e: raise Exception, 'Error aligning second sequence. ' + str( e ) # generate align try: tmp = tempfile.NamedTemporaryFile( dir=tmp_dir ).name tmp_stderr = open( tmp, 'wb' ) proc = subprocess.Popen( args=cmd3, shell=True, cwd=tmp_dir, stderr=tmp_stderr.fileno() ) returncode = proc.wait() tmp_stderr.close() # get stderr, allowing for case where it's very large tmp_stderr = open( tmp, 'rb' ) stderr = '' try: while True: stderr += tmp_stderr.read( buffsize ) if not stderr or len( stderr ) % buffsize != 0: break except OverflowError: pass tmp_stderr.close() if returncode != 0: raise Exception, stderr except Exception, e: raise Exception, 'Error generating alignments. ' + str( e ) # remove header if necessary if options.suppressHeader == 'true': tmp_out = tempfile.NamedTemporaryFile( dir=tmp_dir) tmp_out_name = tmp_out.name tmp_out.close() try: shutil.move( options.output, tmp_out_name ) except Exception, e: raise Exception, 'Error moving output file before removing headers. ' + str( e ) fout = file( options.output, 'w' ) for line in file( tmp_out.name, 'r' ): if not ( line.startswith( '@HD' ) or line.startswith( '@SQ' ) or line.startswith( '@RG' ) or line.startswith( '@PG' ) or line.startswith( '@CO' ) ): fout.write( line ) fout.close() # check that there are results in the output file if os.path.getsize( options.output ) > 0: sys.stdout.write( 'BWA run on %s-end data' % options.genAlignType ) else: raise Exception, 'The output file is empty. You may simply have no matches, or there may be an error with your input file or settings.' except Exception, e: stop_err( 'The alignment failed.\n' + str( e ) ) finally: # clean up temp dir if os.path.exists( tmp_index_dir ): shutil.rmtree( tmp_index_dir ) if os.path.exists( tmp_dir ): shutil.rmtree( tmp_dir ) if __name__=="__main__": __main__()