Notice:
If you happen to see a question you know the answer to, please do chime in and help your fellow community members. We encourage our fourm members to be more involved, jump in and help out your fellow researchers with their questions. GATK forum is a community forum and helping each other with using GATK tools and research is the cornerstone of our success as a genomics research community.We appreciate your help!

Test-drive the GATK tools and Best Practices pipelines on Terra


Check out this blog post to learn how you can get started with GATK and try out the pipelines in preconfigured workspaces (with a user-friendly interface!) without having to install anything.

GATK4 Mutect2 variant with a "germline" flag from unmatched tumors

Hi GATK team,

I have been calling somatic mutations from unmatched tumor samples using the latest GATK4 and found one of the well known EGFR exon deletion variants was flagged as "germline" (please see below) even though it is not present in the germline resource (af-only-gnomad.raw.sites.b37.vcf.gz). I was wondering how GERMQ is calculated.

Thank you,
Jungmin

7 55242465 . GGAATTAAGAGAAGCA G . germline CONTQ=93;DP=58;ECNT=1;GERMQ=1;MBQ=34,35;MFRL=188,206;MMQ=60,60;MPOS=34;POPAF=7.30;RCNTS=0,0;ROQ=93;SEQQ=93;STRANDQ=93;TLOD=70.40 GT:AD:AF:DP:F1R2:F2R1:SB 0/1:36,20:0.362:56:22,13:11,7:21,15,12,8

Best Answer

Answers

  • bshifawbshifaw Member, Broadie, Moderator admin

    Hi @crackiee

    There is a pdf that goes into details about the mathematical aspects of mutect2, here. Try reviewing section E ,Germline Filter, on page 7 for details on how the tool filters germline sites.

    I've only heard of mutect2 being used on matched tumor-normal to help filter out germline sites, how come your using unmatched tumor-normal samples?

  • crackieecrackiee Member

    Hi @bshifaw, this is very useful. My understanding was that mutect2 will use population allele frequency from gnomAD to filter out germline for unmatched samples.
    I will have to go over the germline filter section but it is just counter-intuitive for a variant not seen in gnomAD to get high probability of being called as germline.

  • bshifawbshifaw Member, Broadie, Moderator admin

    Mutect2 rejects from consideration any sites that are most likely to be germline or artifacts based on the paired Normal, a germline population resource, and a Panel of Normals (PoN).

    Perhaps it was seen in either the paired Normal or the Panel of Normals.

  • crackieecrackiee Member

    Right, I checked all of them (it is unmatched so there is no paired normal). This variant is not seen in either a germline resource (gnomAD) or a PoN.

  • bshifawbshifaw Member, Broadie, Moderator admin

    Hi @crackiee

    I've asked one the dev members and they said they can take a look if you provide
    1) the tumor-segmentation file which is an output of CalculateContamination and an optional input of FilterMutectCalls (you may not be running the pipeline this way but it is recommended)
    2) the .filteringStats.tsv output of FilterMutectCalls.

  • crackieecrackiee Member

    Hi @bshifaw

    What would be the best way to provide the files to you or the dev members?

    Thanks so much for looking into this!

  • bshifawbshifaw Member, Broadie, Moderator admin

    You can attach it to a comment here or if its to large you can upload it to our ftp server

  • crackieecrackiee Member

    Hi @bshifaw
    Thanks again and please see attached containing four files.
    If there are additional information that you need, feel free to let me know.

    YUL0183Y1-ID.read-orientation-model.tar.gz
    YUL0183Y1-ID_segments.table
    YUL0183Y1-ID_calculatecontamination.table
    YUL0183Y1-ID.mutect2.contamFiltered_orientFiltered.vcf.gz.filteringStats.tsv

  • crackieecrackiee Member

    Dear @bshifaw

    Just checking in to see if you or the dev member could troubleshoot the issue.

    Thanks!

  • bshifawbshifaw Member, Broadie, Moderator admin
    edited August 15

    Hi @crackiee ,

    The dev member has had limited availability due to the recent gatk release along with other projects but expect a response next week.

    Thanks

  • crackieecrackiee Member
    edited August 15
  • davidbendavidben BostonMember, Broadie, Dev ✭✭✭

    @cracklee In tumor-only calling, most variants with allele fraction near 1/2 are germline events. This is true even for variants missing from gnomAD simply because rare germline variants are still more prevalent than somatic variants. Mutect2 attempts to call the optimal balance of sensitivity and precision. In your sample, there are enough somatic calls with low allele fraction that Mutect2 is satisfied with the sensitivity it obtains from those calls and is less willing to gamble on calls with higher allele fraction. If you want to call everything not in gnomAD as somatic (and remember, most of these calls are germline) you can set -default-af 0.

  • crackieecrackiee Member

    @davidben Thanks for your answer. That makes sense for many heterozygous calls but apparently, this might be a rare exception since this is a very well known EGFR exon 19 deletion which is very unlikely to be a germline considering its deleteriousness. It is more likely present as a homozygous variant (LOH) but appears to be heterozygous due to low tumor purity. But I understand Mutect2 tries to find the optimal balance of sensitivity and precision.

  • crackieecrackiee Member

    @davidben Great idea. I have not tried Funcotator yet but will do. Thank you very much again!

Sign In or Register to comment.