Decrease in the read depth of the sample after Mutect2

Question

[kamanovae]

Hi!

I'm trying to find mutations in tumor samples. Sample was sequencing on a gene panel. Everything is going well in my pipeline up to the Mutect2 stage. Before starting the Mutect2, my sample has very good depth coverage. But after Mutect2 in vcf file there are sap with DP=3 or less

I check the found SNP in the program and confirm that in the initial bam this position has a coverage of more than 20

There is my code in snakemake format:

    rule gatk_Mutect2:
            input: bed="out_intersect_bed/{bed}.bed", ref="hg38/ref.fa", bam="srr11097713_1_base_recalibrate.bam", gnomad="af-only-gnomad.hg38.vcf.gz"
            output: vcf="out_Mutect2/srr11097713.mutect2.{bed}.vcf",  f1r2="out_Mutect2/srr11097713.{bed}.f1r2.tar.gz"
            shell: "gatk-4.2.4.1/gatk --java-options ""-Xmx7680m""  Mutect2 -L {input.bed} -R {input.ref} -I {input.bam} -tumor TUMOR --germline-resource {input.gnomad} -O {output.vcf}
 --f1r2-tar-gz {output.f1r2} --create-output-bam-index false

VCF looks like this:

##filtering_status=Warning: unfiltered Mutect 2 calls.  Please run FilterMutectCalls to remove false positives.
##source=Mutect2
##tumor_sample=TUMOR
#CHROM    POS ID  REF ALT QUAL    FILTER  INFO    FORMAT  TUMOR
chr3  37025867    .   G   A   .   .   AS_SB_TABLE=0,0|0,0;DP=1;ECNT=1;MBQ=0,29;MFRL=0,283;MMQ=60,60;MPOS=10;POPAF=7.30;TLOD=3.08  GT:AD:AF:DP:F1R2:F2R1:FAD:SB    0/1:0,1:0.667:1:0,0:0,1:0,1:0,0,1,0

[iraun]

Mutect2 uses local assembly and realignment to detect SNVs and indels, so it is probable that the read counts differ from the original bam file. If I remember correctly it is possible to use -bamout option to output the BAM file that Mutect2 uses internally to determine the haplotypes. Maybe you could use this option and inspect this BAM in IGV.

[kamanovae]

Thanks for the advise!

I got an intermediate Mutect2 bam file. For the SNP that I mentioned in my question, the number of reads is 3. And not alone as it is written in the vsf file (DP=1). However, I don't understand why the Mutect2 realignment the bam file like this and removed a lot of reads. I don't use any quality filters