Genotype Likelihood In 1000 Genomes Vcf Data

Question

[jjc]

I am trying to understand the per sample genotype data given in the following file from 1000 genomes:

ALL.chr7.phase1_release_v3.20101123.snps_indels_svs.genotypes.vcf

I start by explaining what I see, and then I have two questions at the end.

What I see is a FORMAT column with the entry:

GT:DS:GL

with GT and GL defined as in the VCF definition, with DS being a custom format defined as:

##FORMAT=<ID=DS,Number=1,Type=Float,Description="Genotype dosage from MaCH/Thunder">

One of the several things that I am not understanding is how the genotype call is being made in some cases.

For example, there are many entries in which the GL values are -0.48,-0.48,-0.48

My understanding is that then the read data does not allow a call between the reference or the variant allele at such a bialletic.

However, the DS reading seems to be being used. Specifically, looking across the whole file, I run the following awk script:

BEGIN { FS="\t" }
$1=="#CHROM"{ for (i=10;i<=NF;i++) {
                subject[i]=$i
              }
              next
            }
{ for (i=10;i<=NF;i++) {
     if ($i ~ /:-0.48,-0.48,-0.48/) {
         if ( done[$i] == 0 ) { 
             done[$i] = 1 
             print $i , NR, subject[i]
         } 
      }
  }
}

This is executed:

gunzip -c ALL.chr7.phase1_release_v3.20101123.snps_indels_svs.genotypes.vcf.gz | 
         awk -f gl.awk

Shows DS values from 0.000 to 2.000 with granularity of 0.050. As far as I can tell this value, and some other data, seems to allow a genotype call to be made.

What I see, is that if the DS value lies between 0.000 and 0.500 then the phased genotype is called as 0|0, e,g.

0|0:0.150:-0.48,-0.48,-0.48     39     HG00140

If the DS value lies between 0.550 and 1.000 then the genotype is called as one of 0|0, 1|0, or 0|1 e.g

1|0:0.850:-0.48,-0.48,-0.48    34     HG01168
0|0:0.850:-0.48,-0.48,-0.48    34     HG00242
0|1:0.850:-0.48,-0.48,-0.48    34     HG00143

If the DS value is between 1.050 and 1.500 then the genotype is called as one of 1|1, 1|0, or 0|1 e.g.

0|1:1.200:-0.48,-0.48,-0.48 31 NA19108
1|1:1.200:-0.48,-0.48,-0.48 32 HG01069
1|0:1.200:-0.48,-0.48,-0.48 34 NA12383

If the DS value is between 1.550 and 2.000 then the genotype is called as 1|1.

My two questions are:

1. When the DS lies between 0.5 and 1.5 and it does not indicate definitively either 0|0 or 1|1 on what basis is the genotype call made, and by which piece of software?

2. What is this DS value anyway, and how does it allow the 0|0 and 1|1 calls to be made? "Genotype dosage from MaCH/Thunder" is somewhat cryptic.

Answer 1

[Dan]

Looking here: http://genome.sph.umich.edu/wiki/MaCH

Seems like this is related to imputation, i.e. there may not be direct evidence for the call, but a call is imputed. Sorry I can't be more help... Perhaps you can try asking your question on the VCF mailing list?

VCF format discussions, such as clarifications or proposed changes to the spec: https://lists.sourceforge.net/lists/listinfo/vcftools-spec

Not sure if there is a 1000 genomes mailing list.

HTH, Dan.

[Laura]

This is correct, as the 1000genomes is low coverage all positions do not share equal coverage in all individuals so imputation and haplotype inference is needed to give every individual a genotype are every position

Dosage isn't necessarily 0, 1 or 2 but can be a continum as it shows how likely the genotype is on the basis of the imputation evidencehttp://www.1000genomes.org/faq/what-does-genotype-dosage-mean-phase1-integrated-call-set

[jjc]

Thanks, this does answer the first of my two questions. The calls appear to be made on the basis of haplotype information