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.

m2 gatk4 oncotated maf output includes rejected mutations

amaroamaro Broad InstituteMember, Broadie
edited February 2018 in Ask the GATK team

The oncotated maf output includes many rejected mutations (using the configuration from the public spaces). This is bad practice. The unfiltered VCF (or preferably a tsv) should have these sites but we should not be annotating them or putting them in final outputs.

Issue · Github
by shlee

Issue Number
4421
State
closed
Last Updated
Assignee
Array
Closed By
sooheelee

Answers

  • shleeshlee CambridgeMember, Broadie ✭✭✭✭✭

    Hi @amaro,

    I've moved your question to the GATK forum and I will try to help you as much as I can.

    First, can you confirm which workflow you are using? We need to know which versions of tools or scripts are being run so we can correct for problems where they arise. Is the workflow by chance the recently published: https://portal.firecloud.org/#workspaces/help-gatk/Somatic-SNVs-Indels-GATK4?

    If so, and the WDL script corresponds to GATK4.0.1.1's mutect2.wdl, then I believe there exists NO step that removes filtered variants from the callset. The resulting VCF format callset will contain all calls and be annotated in the FILTER column (column 7) with PASS or filter reason. If this or another iteration of this pipeline is what you are using, then I would ask @LeeTL1220 to jump in here.

    I see a number of choices here to help you get to where you want and you should let us know if any of these will enable you to move forward or if you have another solution in mind that we can implement that will help you.

    1. Move Oncotator/Functotator annotation tasks to their own stand-alone WDL script. This enables (a) manual review and amendment of the FilterMutectCalls/FilterByOrientationBias M2 callset and (b) preparation of a VCF containing only PASS calls.
    2. Add a SelectVariants or awk '$7==PASS"' step to the current script to retain only passing calls that then get passed to the annotation task.
    3. Wait for then use Funcotator. The next release should have an option to ignore filtered sites and support MAF outputs. The earliest these updates will be merged to master is early next week. I believe then either there may be a minor release OR you should be able to use a nightly release GCR docker image.
  • LeeTL1220LeeTL1220 Arlington, MAMember, Broadie, Dev ✭✭✭

    @amaro We were (twice) originally asked to keep the filtered mutations, but you are not the first to request this change. Currently, Oncotator (1.9.6.1) cannot remove the filtered mutations.
    Another suggestion, if you need this immediately:
    -- If you are able to modify the WDL, you can specify --collapse-filter-cols in the Oncotator command line. This will collapse the filter columns to one and then egrep/awk/etc should be pretty simple.

    Soo Hee's suggestion in #2 would also work well.

    I'll try to get out a fix soon.

  • LeeTL1220LeeTL1220 Arlington, MAMember, Broadie, Dev ✭✭✭

    @amaro I cut a new release. Use the oncotator_docker broadinstitute/oncotator:1.9.7.0 and you will have to update the M2 oncotator WDL to add the flags --collapse-filter-cols and --prune-filter-cols.

    At some point soon, we will have a mechanism to turn this on as a M2 WDL parameter -- please see: https://github.com/broadinstitute/gatk/issues/4422

    For now, you will have to modify the WDL.

  • LeeTL1220LeeTL1220 Arlington, MAMember, Broadie, Dev ✭✭✭
    edited February 2018

    @amaro Just to be clear, oncotator 1.9.7.0 does the filtering of variants on your behalf. Variants filtered in the VCF will not be in your MAF if you follow the instructions above. You just need to upgrade your docker image and alter your WDL (until we release a new Mutect2 WDL that turns this on).

Sign In or Register to comment.