We've moved!
This site is now read-only. You can find our new documentation site and support forum for posting questions here.
Be sure to read our welcome blog!

GATK Mutect2 doesn't emit a real variant

amjaddamjadd FinlandMember ✭✭

Hello GATK team,

I have been trying the brand new Mutect2 v4.1.1.0 on ctDNA samples with high coverage. I noticed one case where Mutect2 v4.1.0.0 detected a variant, but the new version didn't even emit the mismatch.

Here is the original command I used:

gatk Mutect2 -R reference.fasta -I normal.bam \
-O variants.vcf.gz --germline-resource gnomadhg19.vcf.gz \
--panel-of-normals pon.vcf.gz -L targets.interval_list \
-ip 300 -normal normalSample -I tumor1.bam \
-I tumor2.bam -I tumor3.bam -I tumor4.bam

When I realized the missing variant, I reran Mutect2 with the following additional parameters, but that didn't help.

--force-active true --tumor-lod-to-emit 0 --initial-tumor-lod 0

Here is an IGV snapshot of the variant (it exists in only one tumor sample: Ref: 820, Alt: 29, N: 2)

The reads in the upper part are from the original bam and in the lower part from the bam file emitted by Mutect2

Thank you for your help


Best Answer


  • amjaddamjadd FinlandMember ✭✭

    Update: The variant was detected after setting --max-reads-per-alignment-start 0.
    And now the reported counts better represent the depth of the samples.
    I'm looking forward to seeing the blog about ctDNA and parameter recommendations for high-depth samples.

  • bhanuGandhambhanuGandham Cambridge MAMember, Administrator, Broadie, Moderator admin


    I am glad you were able to figure it out. I have asked the dev for an explanation of why that happens and will get back to you shortly.

  • amjaddamjadd FinlandMember ✭✭

    Thank you for answering @bhanuGandham
    All clear

Sign In or Register to comment.