Problems with callers of TCGA data

Question

[jaumesastretomas]

We have restricted access to the TCGA data portal and we downloaded VCF files from LUAD project.

We are facing 2 problems with the files that we don't understand:

VARSCAN2 VCF files annotate normal sample always like 0/0, why is that? Do not call the normal variants?

MUTECT VCF files annotate tumor samples always like heterozygous 0/1. Why is that? Aren't there homozygous variants?

Thank you very much for your attention.

Answer 1

[Kevin Blighe]

I cannot give a definitive answer, but 0/0 in VCF format just means that the variant doesn't exist in the sample. I typically see 0/0 when I merge multiple samples into a single VCF. The VCF has to list each variant 1-per-row, so, many variants will exist in one sample but not others.

Regarding the heterozygous calls, I can only imagine that a very heterogeneous tumour bulk sample was sequenced. A heterogeneous bulk sample will contain many tumour clones (thousands or millions - who knows); thus, the frequencies of the variants would be diluted based on the number of different clones. Depending on the cut-offs used by MUTECT, a homozygous variant may never be called in such a scenario. I don't fully know how MUTECT functions, but this issue appears to have been reported on the Broad Institute's forum: https://gatkforums.broadinstitute.org/gatk/discussion/6865/mutect-unable-to-detect-some-germline-mutations

[Brian Bushnell]

I cannot give a definitive answer, but 0/0 in VCF format just means that the variant doesn't exist in the sample.

Are you sure about that? I use it to indicate that the variant does not represent the majority of either the paternal or maternal chromosome copies in a diploid, though it may still exist as a somatic variation.

[Kevin Blighe]

...fairly sure, but you've now cast some doubt in my mind! :/