Hard filtering on GATK HaplotypeCaller giving multiple warnings

Question

[graeme.thorn]

I'm using this pipeline for deriving variants from RNA sequencing data: https://github.com/modupeore/VAP which uses specific versions of various tools, including HaplotypeCaller from GATK (v3.8-0-ge9d806836). The final step is a set of hard filters on the called variants (applied using VariantFilter), but looking at the log files, there are a lot of lines of the following:

WARN  03:22:38,287 Interpreter - ![0,2]: 'QD < 5.0;' undefined variable QD
WARN  03:22:38,288 Interpreter - ![0,14]: 'ReadPosRankSum < -8.0;' undefined variable ReadPosRankSum
WARN  03:22:38,288 Interpreter - ![0,2]: 'MQ < 40.0;' undefined variable MQ
WARN  03:22:38,288 Interpreter - ![0,9]: 'MQRankSum < -12.5;' undefined variable MQRankSum
WARN  03:22:38,288 Interpreter - ![0,2]: 'FS > 60.0;' undefined variable FS

The vcf generated by HaplotypeCaller has multiple lines where these are not present in the INFO string, so the filters are not applied, and these are warnings so can be ignored, to some extent. Is there a way to force HaplotypeCaller to output these calculations for every line so these filters can be applied, or is it worth using the latest version (and forcing that to output these INFO values) instead? (The maintainer of the pipeline recommends not to upgrade, for an unknown reason)

Answer 1

[graeme.thorn]

It turns out, via reading the GATK forums, that HaplotypeCaller doesn't generate these variables in INFO if the site is homozygous alternate, and there are no reference reads: https://gatk.broadinstitute.org/hc/en-us/community/posts/360072160371-Whether-to-filter-rows-without-ReadPosRankSum-and-MQRankSum-

As such, the warnings could probably be discarded, or a different set of filters applied that doesn't include these variables.