improve the LD statistics

[samtools.git] / bcftools / vcfutils.pl

diff --git a/bcftools/vcfutils.pl b/bcftools/vcfutils.pl
index 6fb1952cc6e578524078d6dcef12b65cd760a4ed..d0b797143cf56fe0a67383ec4f610096eaecf2c0 100755 (executable)
--- a/bcftools/vcfutils.pl
+++ b/bcftools/vcfutils.pl
@@ -14,7 +14,7 @@ sub main {
   &usage if (@ARGV < 1);
   my $command = shift(@ARGV);
   my %func = (subsam=>\&subsam, listsam=>\&listsam, fillac=>\&fillac, qstats=>\&qstats, varFilter=>\&varFilter,
-                         hapmap2vcf=>\&hapmap2vcf, ucscsnp2vcf=>\&ucscsnp2vcf);
+                         hapmap2vcf=>\&hapmap2vcf, ucscsnp2vcf=>\&ucscsnp2vcf, filter4vcf=>\&filter4vcf, ldstats=>\&ldstats);
   die("Unknown command \"$command\".\n") if (!defined($func{$command}));
   &{$func{$command}};
 }
@@ -70,12 +70,21 @@ sub fillac {
          print;
        } else {
          my @t = split;
-         my @c;
+         my @c = (0);
          my $n = 0;
-         $c[1] = 0;
+         my $s = -1;
+         @_ = split(":", $t[8]);
+         for (0 .. $#_) {
+               if ($_[$_] eq 'GT') { $s = $_; last; }
+         }
+         if ($s < 0) {
+               print join("\t", @t), "\n";
+               next;
+         }
          for (9 .. $#t) {
-               if ($t[$_] =~ /^(\d+).(\d+)/) {
-                 ++$c[$1]; ++$c[$2];
+               if ($t[$_] =~ /^0,0,0/) {
+               } elsif ($t[$_] =~ /^([^\s:]+:){$s}(\d+).(\d+)/) {
+                 ++$c[$2]; ++$c[$3];
                  $n += 2;
                }
          }
@@ -94,8 +103,30 @@ sub fillac {
   }
 }
 
+sub ldstats {
+  my %opts = (t=>0.9);
+  getopts('t:', \%opts);
+  die("Usage: vcfutils.pl ldstats [-t $opts{t}] <in.vcf>\n") if (@ARGV == 0 && -t STDIN);
+  my $cutoff = $opts{t};
+  my ($last, $lastchr) = (0x7fffffff, '');
+  my ($x, $y, $n) = (0, 0, 0);
+  while (<>) {
+       if (/^([^#\s]+)\s(\d+)/) {
+         my ($chr, $pos) = ($1, $2);
+         if (/NEIR=([\d\.]+)/) {
+               ++$n;
+               ++$y, $x += $pos - $last if ($lastchr eq $chr && $pos > $last && $1 > $cutoff);
+         }
+         $last = $pos; $lastchr = $chr;
+       }
+  }
+  print "Number of SNP intervals in strong LD (r > $opts{t}): $y\n";
+  print "Fraction: ", $y/$n, "\n";
+  print "Length: $x\n";
+}
+
 sub qstats {
-  my %opts = (r=>'', s=>0.01, v=>undef);
+  my %opts = (r=>'', s=>0.02, v=>undef);
   getopts('r:s:v', \%opts);
   die("Usage: vcfutils.pl qstats [-r ref.vcf] <in.vcf>\n
 Note: This command discards indels. Output: QUAL #non-indel #SNPs #transitions #joint ts/tv #joint/#ref #joint/#non-indel \n") if (@ARGV == 0 && -t STDIN);
@@ -147,14 +178,20 @@ Note: This command discards indels. Output: QUAL #non-indel #SNPs #transitions #
   my $next = $opts{s};
   my $last = $a[0];
   my @c = (0, 0, 0, 0);
+  my @lc;
+  $lc[1] = $lc[2] = 0;
   for my $p (@a) {
        if ($p->[0] == -1 || ($p->[0] != $last && $c[0]/@a > $next)) {
          my @x;
          $x[0] = sprintf("%.4f", $c[1]-$c[2]? $c[2] / ($c[1] - $c[2]) : 100);
          $x[1] = sprintf("%.4f", $hsize? $c[3] / $hsize : 0);
          $x[2] = sprintf("%.4f", $c[3] / $c[1]);
+         my $a = $c[1] - $lc[1];
+         my $b = $c[2] - $lc[2];
+         $x[3] = sprintf("%.4f", $a-$b? $b / ($a-$b) : 100);
          print join("\t", $last, @c, @x), "\n";
          $next = $c[0]/@a + $opts{s};
+         $lc[1] = $c[1]; $lc[2] = $c[2];
        }
        ++$c[0]; $c[1] += $p->[1]; $c[2] += $p->[2]; $c[3] += $p->[3];
        $last = $p->[0];
@@ -301,6 +338,45 @@ sub varFilter_aux {
   }
 }
 
+sub filter4vcf {
+  my %opts = (d=>3, D=>2000, 1=>1e-4, 2=>1e-100, 3=>0, 4=>1e-4, Q=>10, q=>3);
+  getopts('d:D:1:2:3:4:Q:q:', \%opts);
+  die(qq/
+Usage:   vcfutils.pl filter4vcf [options] <in.vcf>
+
+Options: -d INT     min total depth (given DP or DP4) [$opts{d}]
+         -D INT     max total depth [$opts{D}]
+         -q INT     min SNP quality [$opts{q}]
+         -Q INT     min RMS mapQ (given MQ) [$opts{Q}]
+         -1 FLOAT   min P-value for strand bias (given PV4) [$opts{1}]
+         -2 FLOAT   min P-value for baseQ bias [$opts{2}]
+         -3 FLOAT   min P-value for mapQ bias [$opts{3}]
+         -4 FLOAT   min P-value for end distance bias [$opts{4}]\n
+/) if (@ARGV == 0 && -t STDIN);
+
+  my %ts = (AG=>1, GA=>1, CT=>1, TC=>1);
+
+  my @n = (0, 0);
+  while (<>) {
+       if (/^#/) {
+         print;
+         next;
+       }
+       next if (/PV4=([^,]+),([^,]+),([^,]+),([^,;\t]+)/ && ($1<$opts{1} || $2<$opts{2} || $3<$opts{3} || $4<$opts{4}));
+       my $depth = -1;
+       $depth = $1 if (/DP=(\d+)/);
+       $depth = $1+$2+$3+$4 if (/DP4=(\d+),(\d+),(\d+),(\d+)/);
+       next if ($depth > 0 && ($depth < $opts{d} || $depth > $opts{D}));
+       next if (/MQ=(\d+)/ && $1 < $opts{Q});
+       my @t = split;
+       next if ($t[5] >= 0 && $t[5] < $opts{q});
+       ++$n[0];
+       my @s = split(',', $t[4]);
+       ++$n[1] if ($ts{$t[3].$s[0]});
+       print;
+  }
+}
+
 sub ucscsnp2vcf {
   die("Usage: vcfutils.pl <in.ucsc.snp>\n") if (@ARGV == 0 && -t STDIN);
   print "##fileformat=VCFv4.0\n";
@@ -394,6 +470,7 @@ Command: subsam       get a subset of samples
          fillac       fill the allele count field
          qstats       SNP stats stratified by QUAL
          varFilter    filtering short variants
+         filter4vcf   filtering VCFs produced by samtools+bcftools
          hapmap2vcf   convert the hapmap format to VCF
          ucscsnp2vcf  convert UCSC SNP SQL dump to VCF
 \n/);