VCF Indel line structure: what is actually reported?

Question

[Matteo Schiavinato]

Hi folks,

I am digging in the deepest of variant calling this year, and I stepped on a weird case on my VCF file. INDEL lines usually report reference position and reference allele indicating the nucleotide which is before the indel.

Example: If I have an AAG insertion at position 5 of my scaffold, I will get reported a VCF line like:

Chrom       Pos Tag Ref Alt 
Scaffold    4   .   G   GAAG     ...(etc)

What happens in my file is:

Chrom       Pos Tag Ref Alt 
Scaffold    4   .   TAG TAGAAG     ...(etc)

Not that in the second case the "TA" before the "G" are also included. I checked and these bases are part of the reference and part of 95% of the reads that map there, same reads that call the subsequent indel.

What is happening? Why is bcftools call reporting also those ones into the reference and alternative allele of the indel?

[Santosh Anand]

Is this position multiallelic (at least a SNP?). Which version of VCF?

[Matteo Schiavinato]

VCF 4.2, position is not multiallelic.

Answer 1

[Pierre Lindenbaum]

Depends of your version of samtools/bcftools but AFAIK, the INDEL are sometimes poorly reported. There are some tools to left align the variants:

• https://software.broadinstitute.org/gatk/gatkdocs/3.6-0/org_broadinstitute_gatk_tools_walkers_variantutils_LeftAlignAndTrimVariants.php
• https://github.com/quinlan-lab/vcftidy
• ...

[Matteo Schiavinato]

I'm making my tool myself, hence this question haha. Thank you for the links! Will check.