VCF deletions incorrectly formatted

Question

[john.michel.rouhana]

Hi all,

I'm working with a vcf (v4.1) that has incorrectly formatted deletions for some reason. The insertions are fine, but the deletions are annotated as (example):

#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO    FORMAT
2   32474671    indel.60227 A   -   .   PASS    .   GT

Notice that the ALT is -, when the line should have been formatted as such (example):

#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO    FORMAT
2   32474670    indel.60227 GA  G   .   PASS    .   GT

I have no idea how the deletions ended up like this in the vcf, but my present plan is to parse a reference genome fasta file for these positions and manually correct all the deletion annotations, so I don't have to drop them from the vcf. What I wanted to know is if there's a tool that already does this- as it stands, I'm writing a manual parser.

[Ram]

It is quite odd that insertions are fine and deletions are not. Older VCF versions (4.1) had . for REF in insertions and . for ALT in deletions, so either both should be affected or neither should be.

Maybe give this tool a shot? Disclaimer: This tool is not mine and I have never used it. Maybe bcftools norm --check-ref can fix the REF alleles, I'm not sure though.

[john.michel.rouhana]

I definitely agree that it's odd. I'm having trouble finding older versions that used a - as ALT in deletions, so I'm not sure it's ever the case. A big part of this problem is that I can't figure out how the people who supplied the VCF ended up in this situation.

bcftools doesn't seem to fix the problem, probably because the REF alleles are fine; it's the ALT that are botched.

Looking into the other tool that you linked. Hopefully it helps.

[Ram]

I don't know of any tool that uses - - older versions used ., not -.

Answer 1

[john.michel.rouhana]

I wound up just parsing the vcf in Python and calling samtools to correct the deletions in the vcf. By subtracting 1 from the alleged bp_pos, calling samtools faidx <reference_genome> chr:del_start-del_end for each REF, and taking the first character of that for ALT, you can fill in the blanks. I don't believe there's a tool that corrects this problem in VCFs, because I don't think this is a common (or normal) problem. Leaving this here in case anyone ever encounters the same situation.