Missing alleles for a genotype in UYG VCF file

Question

[Neilfws]

I'm trying to use Beagle 4.1 to phase a VCF file from Illumina's UYG program, using 1000 Genomes phase 3 as a reference panel. It's failing with the error:

ERROR: Missing one or both alleles for a genotype:

Indeed, when I examine the VCF file I see lines like this one (scroll right):

chr12   1899470 .       C       T       239     PASS    SNVSB=-26.8;SNVHPOL=3;CSQ=T||NM_172364.4|Transcript|downstream_gene_variant|||||||||CACNA2D4|||||1653|YES||||NP_758952.4|||||,T||NM_024551.2|Transcript|downstream_gene_variant|||||||||ADIPOR2|||||1625|YES||||NP_078827.2|||||        GT:GQ:GQX:DP:DPF:AD     1:33:33:18:2:0,17

where the value for GT = 1.

Questions:

• is GT = 1 valid VCF? I had the impression it was not
• is there a smart way to make Beagle ignore these lines? I didn't see anything in the documentation
• or a way to remove these lines in preprocessing using e.g. vcftools?

[Pierre Lindenbaum]

a way to remove these lines:

using vcffilterjs : https://github.com/lindenb/jvarkit/wiki/VCFFilterJS remove the lines having one genotype where num(alleles)!=2.

$ java -jar jvarkit/dist/vcffilterjs.jar -e 'function accept(v) { var i;for(i=0;i< v.getNSamples();i++) { var g=v.getGenotype(i); if(g.isCalled() && g.getAlleles().size()!=2) return false;} return true;} accept(variant);' input.vcf > out.vcf

Answer 1

[Jorge Amigo]

well that's strange... anyway, you can remove "1" genotypes excluding them with bcftools

bcftools view -e 'GT="1"' file.vcf.gz