Somatic variants in different tumor states with no healthy control samples

Question

[jaime alvarez]

Hello:

I have multiple tumor samples with no matched healthy control. For each tumor sample I have 3 states: the original tumor tissue and two other states (in which I put the original tumor tissue and want to find variation with respect to the original tumour) which I will call state2 and state3.

I have tried using GATK germline pipeline for each state compared to the reference genome and get some variants, but I think I am only getting variants which are very represented such as driver mutations.

I have two questions to answer:

1. Variation of the original tumor vs. reference genome.
2. Variation of state2 vs. original tumor and variation of state3 vs. original tumor

For the first question I was thinking in using GATK mutect with no matched control but I think I will get a lot of false positives, so I am considering using software such as the one from the following publications: : https://pubmed.ncbi.nlm.nih.gov/34632388/ https://pubmed.ncbi.nlm.nih.gov/36699358/ https://pubmed.ncbi.nlm.nih.gov/36637201/

For the second question, I wanted to ask if it is possible to use GATK mutect (or any general software) with the matched "control" being the original tumor tissue and each of the states (state2 and state3) being the "tumour" sample.

Thank you.

Answer 1

[Ram]

MuTect2 allows for tumor-only calling using a 1000genomes-based pooled normal as well as a gnomAD-based germline resource. It can give you a few false positives but not a ton of them. It's still worth it.

As for using a tumor state as a pseudo-normal, see #6 on this page: https://gatk.broadinstitute.org/hc/en-us/articles/360050722212-FAQ-for-Mutect2

[jaime alvarez]

Thank you for the reply. In the case of using a tumor state as a pseudo-normal, ¿is it still recommended to use Panel Of Normals (PON)? If affirmative, ¿should they be composed of healthy individuals?

[Ram]

I'm sorry, I cannot help you there as I do not have any experience with that.

Theoretically though, you don't want to include a panel of normals since the set of variants you're looking for are basically tumor-state-2 minus tumor-state-one. Any germline variants will automatically be eliminated as will any somatic variants seen in both tumor states. Again, this is theoretical and not from experience.