high frequency SNPs (and other variations) after subtracting blood variants from tumor ones

Question

[gab]

Hi everybody!

DISCLAIMER: I am new and very glad to join this community, therefore I will gladly accept all answers that pointing out I have done/written something wrong/misplaced/redundant.

As first job in bioinformatics and data analysis, I'm looking for mutations occurring in some tumor samples. I have a number of tumor WES each with corresponding blood WES. When I tried to subtract blood mutations from the ones i got in the tumors, I got some hundreds of remaining variants for each blood-tumor pair. Looking deeper into this, I found out a good half of them has a gnomAD or dbSNP population frequency over 1% or more (max=26%). This is bothering me because I should be left with mostly somatic and pathogenic/rare variations, therefore I expect their frequency in the population to be very low. I ask you if you have any idea why: are my reasoning/expectation wrong? Should I check the quality of sequencing data (although it seems good)? Is this a coverage problem?

Thanks in advance! Gab

Answer 1

[Chris Miller]

If you're trying to call somatic variants, you should use a somatic variant caller (Such as Mutect, Strelka, or VarScan). The results will be different than if you simply compare tumor SNPs to normal SNPs, as there are statistical assumptions that are very different between the two models.