Using uninformative reads (DP) in Mutect2

Follow

James Melhorn

• January 08, 2024 12:20

Dear GATK
I am using Mutect2 to call human somatic variants. Coverage is fairly deep (median gt_AD for the reference allele ~1200 reads). I understand DP as total reads at the site and AD as the 'informative', post filter reads and that therefore DP and gt_AD can be different. However, in some cases the number of DP 'uninformative' reads is of a significant magnitude higher than the gt_AD reads (reference and alternate allele combined) e.g., in the example below around x18K higher. What inferences may I take from this and can I use this discrepancy to filter these variants as very unlikely? 

CHROM

POS

ID

REF

ALT

AS_FilterStatus

FILTER

QUAL

variant_type

allele_type

gt_GT

DP

gt_AD

chr7

1.43E+08

rs1042955

G

A

SITE

PASS

NA

snv

snv

0/1

28736874

1011,532

0

• 

  David Benjamin

  • January 10, 2024 05:02

  This is not the output of Mutect2 / FilterMutectCalls.  If you give us an example as it appears in the VCF output of FilterMutectCalls we might be able to diagnose the issue.  Also, gt_AD is not a GATK annotation so we can't offer an opinion on it.  If there is a discrepancy between DP and AD we can try to explain it.

  0

  Comment actions Permalink

Please sign in to leave a comment.