Mutect2 calling InDel as separate insertion and deletion even after max_mnp_dist is set high

Follow

Yuvanesh V

• July 12, 2024 16:21

a) GATK version used:  4.6.0.0

b) Exact command used: 

    mutect2=".../gatk-4.6.0.0/gatk Mutect2 \

        --java-options -Xmx40G  \

        -R ${ref_genome} \

        -I:tumor ${bam} \

        -L ${bed} \

        --max-reads-per-alignment-start 0 \

        --min-base-quality-score 20 \

        --native-pair-hmm-threads 4 \

        --max-mnp-distance 10 \

        --output output.vcf"

 

I need help to understand why max_mnp_distance is not working.

 

There are 2 variants (insertions and deletions) in a sample which is not merged and called as 1 "InDels"

 

chr9 136496337 . G GTGTGA . . AS_SB_TABLE=466,294|46,24;DP=879;ECNT=2;ECNTH=2;MBQ=20,20;MFRL=148,137;MMQ=60,60;MPOS=26;POPAF=7.30;TLOD=226.04 GT:AD:AF:DP:F1R2:F2R1:FAD:PGT:PID:PS:SB 0|1:760,70:0.094:830:236,15:236,32:539,55:0|1:136496337_G_GTGTGA:136496337:466,294,46,24

chr9 136496338 . C A . . AS_SB_TABLE=466,294|46,24;DP=882;ECNT=2;ECNTH=2;MBQ=20,20;MFRL=148,137;MMQ=60,60;MPOS=28;POPAF=7.30;TLOD=226.03 GT:AD:AF:DP:F1R2:F2R1:FAD:PGT:PID:PS:SB 0|1:760,70:0.094:830:247,14:241,29:539,55:0|1:136496337_G_GTGTGA:136496337:466,294,46,24

 

I used max_mnp_distance to 10, but these 2 SNV are not merging into 1 MNP.

 

VCF output from Mutect2:

 

VCF output from FreeBayes:

 

 

I came across this git (https://github.com/broadinstitute/gatk/issues/8310). But it did not help my case.

 

I had a similar issue for below case:

Mutect2 (with default max_mnp_dist):

 

Mutect2 (max_mnp_dist=3):

This got corrected.

 

Could you please suggest what I am doing wrong !!

 

Thank you,

Yuvanesh

0

• 

  Gökalp Çelik

  • July 13, 2024 08:15

  Hi Yuvanesh V

  There could be multiple reasons. 

  1. David Benjamin's response about those 2 variants not being on the same haplotype. In that case there is no way to combine them into the same MNP. 

  2. Sometimes more mismatches are more costly compared to a single mismatch and an indel. This is due to default Smith-Waterman parameters within Mutect2. Although they can be modified to allow more mismatches within a given region before starting an INDEL your other true variants may fall into an existential crisis due to suboptimal realignment therefore your whole work may become hard to interpret.

  I hope this helps. 

  0

  Comment actions Permalink

Please sign in to leave a comment.