diff pileup_to_vcf.py @ 0:3890f8ba0e4d

Uploaded
author jjohnson
date Mon, 18 Feb 2013 11:32:50 -0500
parents
children bb0ac51f1b02
line wrap: on
line diff
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/pileup_to_vcf.py	Mon Feb 18 11:32:50 2013 -0500
@@ -0,0 +1,263 @@
+#!/usr/bin/env python
+"""
+#
+#------------------------------------------------------------------------------
+#                         University of Minnesota
+#         Copyright 2012, Regents of the University of Minnesota
+#------------------------------------------------------------------------------
+# Author:
+#
+#  James E Johnson
+#  Jesse Erdmann
+#
+#------------------------------------------------------------------------------
+"""
+
+"""
+Generate a VCF file from a samtools pileup
+filtering on read coverage, base call quality, and allele frequency
+
+Pileup Format
+http://samtools.sourceforge.net/pileup.shtml
+Columns:
+chromosome, 1-based coordinate, reference base, the number of reads covering the site, read bases, base qualities
+
+VCF Format
+http://www.1000genomes.org/wiki/Analysis/Variant%20Call%20Format/vcf-variant-call-format-version-41
+The header line names the 8 fixed, mandatory columns. These columns are as follows:
+CHROM POS ID REF ALT QUAL FILTER INFO
+"""
+
+import sys,re,os.path
+import optparse
+from optparse import OptionParser
+
+vcf_header =  """\
+##fileformat=VCFv4.0
+##source=pileup_to_vcf.pyV1.0
+##INFO=<ID=DP,Number=1,Type=Integer,Description=\"Total Depth\">
+##INFO=<ID=$SAF_header,Number=.,Type=Float,Description=\"Specific Allele Frequency\">
+##FILTER=<ID=DP,Description=\"Minimum depth of %s\">
+##FILTER=<ID=$SAF_header,Description=\"Allele frequency of at least %s with base quality minimum %d\">
+#CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO\
+"""
+
+def __main__():
+  #Parse Command Line
+  parser = optparse.OptionParser()
+  # files
+  parser.add_option( '-i', '--input', dest='input', help='The input pileup file (else read from stdin)' )
+  parser.add_option( '-o', '--output', dest='output', help='The output vcf file (else write to stdout)' )
+  # filters
+  parser.add_option( '-c', '--min_coverage', type='int', default='0', dest='min_coverage', help='The minimum read coverage depth to report a variant [default 0]' )
+  parser.add_option( '-r', '--report_depth', type='choice', choices=['all','ref','qual'], default='ref', dest='report_depth', help='Coverage depth as: ref - reads covercovering base, all - reads spanning base, qual - reads with base call above min_base_quality' )
+  parser.add_option( '-b', '--min_base_quality', type='int', default='0', dest='min_base_quality', help='The minimum base quality for a base call to be counted' )
+  parser.add_option( '-f', '--min_allele_freq', type='float', default='.5', dest='min_allele_freq', help='The minimum frequency of an allele for it to be reported (default .5)' )
+  parser.add_option( '-m', '--allow_multiples', action="store_true", dest='allow_multiples', default=False, help='Allow multiple alleles to be reported' )
+  parser.add_option( '-s', '--snps_only', action="store_true", dest='snps_only', default=False, help='Only report SNPs, not indels' )
+  # select columns
+  parser.add_option( '-C', '--chrom_col', type='int', default='1', dest='chrom_col', help='The ordinal position (starting with 1) of the chromosome column' )
+  parser.add_option( '-P', '--pos_col', type='int', default='2', dest='pos_col', help='The ordinal position (starting with 1) of the position column' )
+  parser.add_option( '-R', '--ref_col', type='int', default='3', dest='ref_col', help='The ordinal position (starting with 1) of the reference base  column' )
+  parser.add_option( '-D', '--coverage_col', type='int', default='4', dest='coverage_col', help='The ordinal position (starting with 1) of the read coverage depth column' )
+  parser.add_option( '-B', '--base_call_col', type='int', default='5', dest='base_call_col', help='The ordinal position (starting with 1) of the base call column' )
+  parser.add_option( '-Q', '--base_qual_col', type='int', default='6', dest='base_qual_col', help='The ordinal position (starting with 1) of the base quality column' )
+  # debug
+  parser.add_option( '-d', '--debug', action="store_true", dest='debug', default=False, help='Print Debug messages' )
+  # 
+  (options, args) = parser.parse_args()
+
+  debug = options.debug if options.debug else False
+
+  #Coverage must have at least a total depth of min_coverage to be considered
+  min_coverage = options.min_coverage
+  #Count depth as: ref - reads covercovering base, all - reads spanning base, qual - reads with base call above min_base_quality
+  dp_as = options.report_depth
+  #Base qualities must be at least the min_base_quality value to be counted
+  min_base_quality = options.min_base_quality
+  #Coverage must be more than this pct in support of the variant, e.g. variant support / total coverage > min_allele_freq
+  min_allele_freq = options.min_allele_freq
+  #Allow multiple variants to be reported for a position, min_allele_freq must be less than 0.5 for this to have effect.
+  allow_multiples = options.allow_multiples
+  # Whether to report indels 
+  report_indels = not options.snps_only
+
+  # Check for 1-based column options, and subtract 1 for python array index
+  #Which column (count starting with 1) contains the chromosome (Default assumes 6-col pileup or 12-col filtered pileup)
+  chrom_col = options.chrom_col - 1
+  #Which column (count starting with 1) contains the position (Default assumes 6-col pileup or 12-col filtered pileup)
+  pos_col = options.pos_col - 1
+  #Which column (count starting with 1) contains the reference base (Default assumes 6-col pileup or 12-col filtered pileup)
+  ref_col = options.ref_col - 1
+  #Which column (count starting with 1) contains the total coverage for the position (Default assumes 6-col pileup or 12-col filtered pileup)
+  coverage_col = options.coverage_col - 1
+  #Which column (count starting with 1) contains the base calls for the position (Default assumes 6-col pileup or 12-col filtered pileup)
+  base_call_col = options.base_call_col - 1
+  #Which column (count starting with 1) contains the base call qualities for the position (Default assumes 6-col pileup or 12-col filtered pileup)
+  base_qual_col = options.base_qual_col - 1
+
+  # pileup input 
+  if options.input != None:
+    try:
+      inputPath = os.path.abspath(options.input)
+      inputFile = open(inputPath, 'r')
+    except Exception, e:
+      print >> sys.stderr, "failed: %s" % e
+      exit(2)
+  else:
+    inputFile = sys.stdin
+  # vcf output 
+  if options.output != None:
+    try:
+      outputPath = os.path.abspath(options.output)
+      outputFile = open(outputPath, 'w')
+    except Exception, e:
+      print >> sys.stderr, "failed: %s" % e
+      exit(3)
+  else:
+    outputFile = sys.stdout
+
+  indel_len_pattern = '([1-9][0-9]*)'
+  ref_skip_pattern = '[<>]'
+
+  SAF_header = "SAF";
+
+  print >> outputFile, vcf_header % (min_coverage,min_allele_freq,min_base_quality)
+
+  try:  
+    for linenum,line in enumerate(inputFile):
+      if debug: print >> sys.stderr, "%5d:\t%s" % (linenum,line)
+      fields = line.split('\t')
+      adp = int(fields[coverage_col]) # count of all reads spanning ref base
+      # Quick check for coverage
+      if adp < min_coverage :  
+        continue
+      bases = fields[base_call_col]
+      if dp_as != 'all' :
+        m = re.findall(ref_skip_pattern,bases) 
+        # Quick check for coverage, pileup coverage - reference skips
+        if adp -len(m) < min_coverage:
+          continue
+      quals = fields[base_qual_col]
+      chrom = fields[chrom_col]
+      pos = int(fields[pos_col])
+      ref = fields[ref_col]
+      deletions = dict()
+      max_deletion = 0
+      insertions = dict()
+      snps = {'A':0,'C':0,'G':0,'T':0,'N':0}
+      bi = 0	# bases index
+      qi = 0	# quals index
+      rdp = 0	# read coverage depth
+      qdp = 0	# read coverage depth with base call quality passing minimum
+      variants = dict() # variant -> count
+      while bi < len(bases):
+        base = bases[bi] 
+        if debug: print >> sys.stderr, "%3d\t%s\t%3d\t%s %3d" % (bi,base,qi,quals[qi],ord(quals[qi]) - 33)
+        bi += 1
+        if base in '<>' :	#reference skip (between paired reads)
+          pass;
+        elif base in '.,' : # match reference on forward/reverse strand
+          rdp += 1
+          qual = ord(quals[qi]) -33
+          qi += 1
+          # count match 
+          if qual >= min_base_quality:
+            qdp += 1
+        elif base in 'ACGTNacgtn' : # SNP on forward/reverse strand
+          rdp += 1
+          qual = ord(quals[qi]) -33
+          qi += 1
+          # record snp variant
+          if qual >= min_base_quality:
+            qdp += 1
+            snps[base.upper()] += 1
+        elif base in '+-' : # indel 
+          rdp += 1
+          qdp += 1 # no related quality score, so count it
+          m = re.match(indel_len_pattern,bases[bi:]) 
+          indel_len_str = m.group(0) # get the indel len string
+          bi += len(indel_len_str) # increment the index by the len of the len string
+          indel_len = int(indel_len_str)
+          indel = bases[bi:bi+indel_len].upper() # get the indel bases
+          bi += indel_len # increment the index by the len of the indel 
+          if base == '+':
+            # record insertion variant
+            if indel in insertions: 
+              insertions[indel] += 1
+            else:
+              insertions[indel] = 1
+          elif base == '-': 
+            # record deletion variant
+            max_deletion = max(indel_len,max_deletion)
+            if indel in deletions: 
+              deletions[indel] += 1
+            else:
+              deletions[indel] = 1
+        elif base == '^' : #start of read (followed by read quality char)
+          read_mapping_qual = ord(bases[bi]) - 33 
+          bi += 1
+        elif base == '$' : #end of read
+          pass # no associated quality in either bases or quals
+      dp = rdp if dp_as == 'ref' else qdp if dp_as == 'qual' else adp
+      if debug: print >> sys.stderr, "snps: %s\tins: %s\tdel: %s" % (snps,insertions,deletions)
+      # output any variants that meet the depth coverage requirement
+      vcf_pos = pos
+      vcf_ref = ref
+      suffix = ''
+      alts = []
+      safs = []
+      # If we have any deletions to report, need to modify the ref string
+      if debug: print >> sys.stderr, "max deletions: %s" % deletions
+      if report_indels:
+        for k,v in deletions.items():
+          saf = v/float(dp)
+          if saf >= min_allele_freq:
+            if len(k) > len(suffix):
+              suffix = k
+        vcf_ref = (ref + suffix).upper()
+      if debug: print >> sys.stderr, "snps: %s" % snps
+      for k,v in snps.items():
+        if debug: print >> sys.stderr, "snp: %s %d" % (k,v)
+        saf = v/float(dp)
+        if saf >= min_allele_freq:
+          alts.append(k + suffix)
+          safs.append(saf)
+      if debug: print >> sys.stderr, "insertions: %s" % insertions
+      if report_indels:
+        for k,v in insertions.items():
+          saf = v/float(dp)
+          if saf >= min_allele_freq:
+            alts.append(ref + k + suffix)
+            safs.append(saf)
+        if debug: print >> sys.stderr, "deletions: %s" % deletions
+        for k,v in deletions.items():
+          saf = v/float(dp)
+          if saf >= min_allele_freq:
+            alts.append(vcf_ref[:len(vcf_ref) - len(k)])   # TODO alt will be a substring of vcf_ref,  test this
+            safs.append(saf)
+      if len(alts) > 0:
+        vcf_id = "."
+        vcf_qual = "." 
+        vcf_filter = "PASS"
+        # if not allow_multiples, report only the most freq alt
+        if not allow_multiples and len(alts) > 1:
+          max_saf = max(safs)
+          for i,v in enumerate(safs):
+            if v == max_saf:
+              vcf_alt = alts[i]
+          info = "SAF=%f;DP=%d" % (max_saf,dp)
+        # if allow_multiples
+        else:
+          vcf_alt = ','.join(alts)
+          isafs = []
+          for saf in safs:
+            isafs.append("SAF=%f" % saf)
+          info = "%s;DP=%d" % (','.join(isafs),dp)
+        print >> outputFile,"%s\t%d\t%s\t%s\t%s\t%s\t%s\t%s" % (chrom,vcf_pos,vcf_id,vcf_ref,vcf_alt,vcf_qual,vcf_filter,info)
+  except Exception, e:
+    print >> sys.stderr, "failed: %s" % e
+    exit(1)
+
+if __name__ == "__main__": __main__()
+