htsworkflow/pipelines/qseq2fastq.py

   1 #!/usr/bin/env python
   2 """Convert a collection of qseq or a tar file of qseq files to a fastq file
   3 """
   4 from __future__ import print_function, unicode_literals
   5 from glob import glob
   6 import os
   7 from optparse import OptionParser
   8 import numpy
   9 import sys
  10 import tarfile
  11
  12 from htsworkflow.util.version import version
  13 from htsworkflow.util.conversion import parse_slice
  14 from htsworkflow.pipelines.desplit_fastq import open_output
  15
  16
  17 def main(cmdline=None):
  18     """Command line driver: [None, '-i', 'tarfile', '-o', 'target.fastq']
  19     """
  20     parser = make_parser()
  21     opts, args = parser.parse_args(cmdline)
  22
  23     if opts.version:
  24         print(version())
  25         return 0
  26
  27     if opts.infile is not None:
  28         qseq_generator = tarfile_generator(opts.infile)
  29     elif len(args) > 0:
  30         qseq_generator = file_generator(args)
  31     else:
  32         qseq_generator = [sys.stdin]
  33
  34     if opts.output is not None:
  35         output = open_output(opts.output, opts)
  36     else:
  37         output = sys.stdout
  38
  39     if opts.nopass_output is not None:
  40         nopass_output = open_output(opts.nopass_output, opts)
  41     else:
  42         nopass_output = None
  43
  44     qseq_parser = Qseq2Fastq(qseq_generator, output, nopass_output)
  45     qseq_parser.fastq = not opts.fasta
  46     qseq_parser.flowcell_id = opts.flowcell
  47     qseq_parser.trim = parse_slice(opts.slice)
  48     qseq_parser.reportFilter = opts.pf
  49
  50     qseq_parser.run()
  51
  52
  53 def make_parser():
  54     """Return option parser"""
  55     usage = "%prog: [options] *_qseq.txt\nProduces Phred33 files by default"
  56     parser = OptionParser(usage)
  57     parser.add_option("-a", "--fasta", default=False, action="store_true",
  58                       help="produce fasta files instead of fastq files")
  59     parser.add_option("-f", "--flowcell", default=None,
  60                       help="Set flowcell ID for output file")
  61     parser.add_option("-i", "--infile", default=None,
  62                       help='source tar file (if reading from an archive '\
  63                            'instead of a directory)')
  64     parser.add_option("-o", "--output", default=None,
  65                       help="output fastq file")
  66     parser.add_option("-n", "--nopass-output", default=None,
  67                       help="if provided send files that failed "\
  68                            "illumina filter to a differentfile")
  69     parser.add_option("-s", "--slice",
  70                       help="specify python slice, e.g. 0:75, 0:-1",
  71                       default=None)
  72     parser.add_option("--pf", help="report pass filter flag", default=False,
  73                       action="store_true")
  74     parser.add_option('--gzip', default=False, action='store_true',
  75                       help='gzip output')
  76     parser.add_option('--bzip', default=False, action='store_true',
  77                       help='bzip output')
  78     parser.add_option("--version", default=False, action="store_true",
  79                       help="report software version")
  80
  81     return parser
  82
  83
  84 def file_generator(pattern_list):
  85     """Given a list of glob patterns yield open streams for matching files"""
  86     for pattern in pattern_list:
  87         for filename in glob(pattern):
  88             yield open(filename, "r")
  89
  90
  91 def tarfile_generator(tarfilename):
  92     """Yield open streams for files inside a tarfile"""
  93     archive = tarfile.open(tarfilename, 'r|*')
  94     for tarinfo in archive:
  95         yield archive.extractfile(tarinfo)
  96
  97
  98 class Qseq2Fastq(object):
  99     """
 100     Convert qseq files to fastq (or fasta) files.
 101     """
 102     def __init__(self, sources, pass_destination, nopass_destination=None):
 103         self.sources = sources
 104         self.pass_destination = pass_destination
 105         if nopass_destination is not None:
 106             self.nopass_destination = nopass_destination
 107         else:
 108             self.nopass_destination = pass_destination
 109
 110         self.fastq = True
 111         self.flowcell_id = None
 112         self.trim = slice(None)
 113         self.report_filter = False
 114
 115     def _format_flowcell_id(self):
 116         """
 117         Return formatted flowcell ID
 118         """
 119         if self.flowcell_id is not None:
 120             return self.flowcell_id + "_"
 121         else:
 122             return ""
 123
 124     def run(self):
 125         """Run conversion
 126         (Used to match threading/multiprocessing API)
 127         """
 128         if self.fastq:
 129             header_template = '@'
 130         else:
 131             # fasta case
 132             header_template = '>'
 133         header_template += self._format_flowcell_id() + \
 134                            '%s_%s:%s:%s:%s:%s/%s%s%s'
 135
 136         for qstream in self.sources:
 137             for line in qstream:
 138                 # parse line
 139                 record = line.rstrip().split('\t')
 140                 machine_name = record[0]
 141                 run_number = record[1]
 142                 lane_number = record[2]
 143                 tile = record[3]
 144                 x = record[4]
 145                 y = record[5]
 146                 #index = record[6]
 147                 read = record[7]
 148                 sequence = record[8].replace('.', 'N')
 149                 quality = convert_illumina_quality(record[9])
 150
 151                 # add pass qc filter if we want it
 152                 pass_qc = int(record[10])
 153                 if self.report_filter:
 154                     pass_qc_msg = " pf=%s" % (pass_qc)
 155                 else:
 156                     pass_qc_msg = ""
 157
 158                 header = header_template % ( \
 159                     machine_name,
 160                     run_number,
 161                     lane_number,
 162                     tile,
 163                     x,
 164                     y,
 165                     read,
 166                     pass_qc_msg,
 167                     os.linesep)
 168
 169                 # if we passed the filter write to the "good" file
 170                 if pass_qc:
 171                     destination = self.pass_destination
 172                 else:
 173                     destination = self.nopass_destination
 174
 175                 destination.write(header)
 176                 destination.write(sequence[self.trim])
 177                 destination.write(os.linesep)
 178                 if self.fastq:
 179                     destination.write('+')
 180                     destination.write(os.linesep)
 181                     destination.write(quality[self.trim].tostring())
 182                     destination.write(os.linesep)
 183
 184 def convert_illumina_quality(illumina_quality):
 185     """Convert an Illumina quality score to a Phred ASCII quality score.
 186     """
 187     # Illumina scores are Phred + 64
 188     # Fastq scores are Phread + 33
 189     # the following code grabs the string, converts to short ints and
 190     # subtracts 31 (64-33) to convert between the two score formats.
 191     # The numpy solution is twice as fast as some of my other
 192     # ideas for the conversion.
 193     # sorry about the uglyness in changing from character, to 8-bit int
 194     # and back to a character array
 195     quality = numpy.asarray(illumina_quality, 'c')
 196     quality.dtype = numpy.uint8
 197     quality -= 31
 198      # I'd like to know what the real numpy char type is
 199     quality.dtype = '|S1'
 200     return quality
 201
 202
 203 if __name__ == "__main__":
 204     main()