java.lang.NullPointerException in GenotypeGVCFs

[amizeranschi]

Hi,

I'm running into a java.lang.NullPointerException for some bcbio runs during the GenotypeGVCFs phase.

This is a known issue with GATK versions >= 4.1.5.0, as outlined here: https://gatk.broadinstitute.org/hc/en-us/community/posts/360060283092-GenotypeGVCFs-java-lang-NullPointerException and here: broadinstitute/gatk#6548.

According to the first link above, a solution is to run GenotypeGVCFs with the setting --genomicsdb-use-vcf-codec: https://gatk.broadinstitute.org/hc/en-us/articles/360040509751-GenotypeGVCFs#--genomicsdb-use-vcf-codec

Is there any way of passing this setting to GenotypeGVCFs from within Bcbio?

[roryk]

Thanks so much for figuring out the source of the problem and how to work around it-- you can pass it via the options config for gatk in your YAML file. I also just pushed a fix so we always run with it enabled while they fix the bug.

[amizeranschi]

Alright, thanks a lot for looking into this.

I didn't realize that one could pass tool-specific settings via the gatk options, as well. Thanks for pointing that out.

[roryk]

No worries, not all of the tools support it, so it is hit or miss if it works or not. Thanks a lot for figuring out what was happening, hopefully the GATK team will fix it soon.

[amizeranschi]

@roryk

Check this out for an update on this issue: broadinstitute/gatk#6675.

Starting from release 4.1.8.0 (which came out 2 days ago and is already in bioconda), the VCF codec was made the default and the argument --genomicsdb-use-vcf-codec has been replaced by --genomicsdb-use-bcf-codec.