Mutect2 error when trying to create fragment with no read support

[avalind]

Bug Report

Affected tool(s) or class(es)

Mutect2

Affected version(s)

• GATK 4.1.4.1

Description

When running Mutect2 (from GATK v4.1.4.1) using the following command:

gatk Mutect2 -R [path to grch37-1kg.fa] -I testcase.bam -O pon.vcf

to create a PoN on NovaSeq WGS-data processed through the best practice pipeline (with the BQSR-steps run through the Spark-enabled tools, and bwa mem with -Y flag) I get the following error in multiple regions:

Stacktrace

AFAIK this is related to the new code path introduced in #6240 and seem to be triggered when there are more than 2 reads supporting a fragment but all of them are either duplicate reads or supplemntary/secondary alignments.

Any input is greatly appreciated. I guess a temporary fix is to use the --independent-mates flag (although haven't tried it yet -- how much worse mutation calling performance do one incur when using that flag?)

Steps to reproduce

Use the following small test case .bam-file as input to the command specified above:

Testcase

Expected behavior

Completion of mutect2 without Exception

Actual behavior

Early termination of the mutect2 run due to raising an exception when trying to create a fragment with no read data to back it up.

---

[johnsCheng]

here i change the gatk-4.1.4.1 to gatk-4.1.2.0，then Mutect can work，
I am not sure what the issue involved. But it keeps reporting the error 'need one or two reads to create fragment' before, however, what i input are totally paired reads after markduplication.

[droazen]

@davidbenjamin I've assigned this one to @KevinCLydon just so he can get some experience debugging GATK -- he may consult you at some point on recommended solutions.

[davidbenjamin]

Closed by #6327.