How to filter a VCF on a set of samples having genotypes containing the minor allele?

Question

[William]

Is there a way to filter a VCF (input stream) on variants where at least a specific subset of the samples in the VCF have a heterozygous or homozygous genotype containing the minor allele?

Or the same filter but then for genotypes containing a non-major allele?

I looked at the bcftools and SnpSift filter documentation and I could not find how to do that in one of these tools. Maybe I overlooked the option or combination of options that I should use?

Closest options that I found are in SnpSift are:

isHom( GEN[0] )
isHet( GEN[0] )
isVariant( GEN[0] )
isRef( GEN[0] )

But SnpSift does not have a isMinorAlleleGenotype( GEN[0] ) functionality.

Is there another tool that can do this?

Or am I best of implementing this myself using a VCF library?

Answer 1

[Pierre Lindenbaum]

GATK VariantAnnotator https://software.broadinstitute.org/gatk/gatkdocs/org_broadinstitute_gatk_tools_walkers_annotator_VariantAnnotator.php

with

 --resource:foo listOfMajorAleleles.vcf  -resourceAlleleConcordance

and then use VariantFiltration https://software.broadinstitute.org/gatk/gatkdocs/org_broadinstitute_gatk_tools_walkers_filters_VariantFiltration.php

to keep/excluse those variants that have been flagged in the previous step.

UPDATE:

Using the following script:

var samples=["S1","S2"];

function accept(v)
    {

    var allele2count={};
    var alleles = v.getAlleles();
    for(var i=0;i< alleles.size();++i)
        {
        allele2count[alleles.get(i).getDisplayString()]=0;
        }

    if(allele2count.length==0) return false;

    for(var i=0;i< v.getNSamples();++i)
        {
        var g = v.getGenotype(i);
        if(!g.isCalled()) continue;
        alleles = g.getAlleles();
        for(var j=0;j< alleles.size();++j)
            {
            allele2count[alleles.get(j).getDisplayString()]++;
            }
        }
    var minor=null;
    for(var a in allele2count)
        {
        if(minor == null || allele2count[minor]>allele2count[a])
            {
            minor=a;
            }
        }
    if(minor==null) return false;
    for(var i in samples)
        {
        var g = v.getGenotype(samples[i]);
        if(!g.isCalled()) continue;
        alleles = g.getAlleles();
        for(var j=0;j< alleles.size();++j)
            {
            if(alleles.get(j).getDisplayString().equals(minor)) return true;
            }
        }
    return false;
    }

accept(variant);

and VCFfilterjs: https://github.com/lindenb/jvarkit/wiki/VCFFilterJS

e.g:

 curl -L "https://raw.githubusercontent.com/lindenb/gatk-ui/master/testdata/mutations.vcf" | java -jar  jvarkit-git/dist/vcffilterjs.jar  -f filter.js

[William]

Thank you Pierre but I was hoping to do this filter on the fly, pre-computing the "listOfMajorAleleles.vcf" is not possible in my situation. Also this does not take any specific set of samples in to account.

[Pierre Lindenbaum]

ah , I see. Please refine what is :" where at least a certain set of samples (in the VCF ?) have a genotype (what kind ?) containing the minor allele? ".

[William]

Updated question description, hope it's more clear now.