I have many VCFs from different samples. If I merge these into a single VCF using vcftools (vcf-merge), the samples where a variant wasn't called are labeled as missing that variant. Instead, I want the VCF to show that the sample has the reference allele (safe to assume in my application).
Is there a way to call missing variants in a VCF as the reference allele? What tools can I use to do this?
EDIT:
The sequences were originally variant called using FreeBayes (through the LongRanger pipeline).
RE-EDIT:
Turns out I can simply use the --ref-for-missing flag in vcf-merge to achieve this. Problem solved.
RE-RE-EDIT:
Using --ref-for-missing flag in vcf-merge does of course not give the variants any annotation, like depth and genotype quality.
( The best way is to call all the BAMs in the same command, to get a multi-sample VCF)
I've written two tools related to your question:
VcfNoCallToHomRef http://lindenb.github.io/jvarkit/VcfNoCallToHomRef.html will arbitrarily convert the ./. to 0/0 .
FixVcfMissingGenotypes http://lindenb.github.io/jvarkit/FixVcfMissingGenotypes.html : will re-use the BAM to test the depth at each NO-CALL site. It's slow;
I think vcf-merge with the --ref-for-missing flag should solve the problem your VcfNoCallToHomRef solves. Your FixVcfMissingGenotypes tool sounds really useful though.
I didn't try your solution because annotating the depth only isn't enough for me. I accepted the answer anyway, as it solves the problem I stated.
Log in to answer this question.
