Can Mutect2 be applied in tumor only mode for somatic variant detection?

sboylesboyle Bay Area, CAMember

I see in your documentation for Mutect2 you state: "MuTect2 currently only supports the calling of a single tumor-normal pair at a time"

However, it appears to work in tumor only mode as well, at least for the creation of the panel of normals.

My question is can Mutect2 be applied in tumor only mode (simply with no normal provided and not using the artifact_detection_mode flag) to detect accurate somatic variants when a normal sample is not available (similarly to what could be done in mutect1)?

Best Answer


  • bracha25bracha25 Member

    Do you have the same recommendation to MuTect ?

  • Geraldine_VdAuweraGeraldine_VdAuwera Cambridge, MAMember, Administrator, Broadie admin

    Yes, the original MuTect has the same limitation. I'm not aware of a somatic variant caller that does a really good job running on tumor only. The role of the normal sample is not just to eliminate germline variants, but also to provide a baseline for estimating artifacts. Having a really good, technically appropriate Panel of Normals can help compensate, but it's not as good as running with a matched normal.

  • ac67479ac67479 AustinMember

    Hi !

    I am in this exact same case where I have no matched normal. But I have data coming from 2 sources : RNAseq and exome sequencing from the same individual. Do you think that finding variants in both samples for the same individuals can give sufficient confidence? We end up with TLOD but there is no NLOD because there is no matched normal. Is the TLOD informative about the likelihood of having a somatic mutation or is the NLOD score the only feature that is informative regarding that point ?

    Issue · Github
    by Sheila

    Issue Number
    Last Updated
    Closed By
  • Geraldine_VdAuweraGeraldine_VdAuwera Cambridge, MAMember, Administrator, Broadie admin

    Hi @c67479, we're going to write up a doc in the near future to explain how TLOD and NLOD are used in MuTect's decision process. In a nutshell, TLOD is somewhat informative in itself since it is the log odds of the site being alternate over being reference in the tumor sample, meaning how confident we are that we've observed an ALT allele -- but it is limited since it does not tell you anything about whether this might just be a variant that is present in the germline background.

    Similarly, while you can reasonably use the presence of a candidate mutation in two separate sources to boost your confidence that there's something there (as opposed to a sequencing artifact) it sill doesn't allow you to rule out that it's a germline variant. For that you can use databases of known variation, but be aware that this won't exclude variants that are private to your sample.

Sign In or Register to comment.