If you happen to see a question you know the answer to, please do chime in and help your fellow community members. We encourage our fourm members to be more involved, jump in and help out your fellow researchers with their questions. GATK forum is a community forum and helping each other with using GATK tools and research is the cornerstone of our success as a genomics research community.We appreciate your help!
Test-drive the GATK tools and Best Practices pipelines on Terra
Check out this blog post to learn how you can get started with GATK and try out the pipelines in preconfigured workspaces (with a user-friendly interface!) without having to install anything.
I'm curious about the experience of the community at large with VQSR, and specifically with which sets of annotations people have found to work well. The GATK team's recommendations are valuable, but my impression is that they have fairly homogenous data types - I'd like to know if anyone has found it useful to deviate from their recommendations.
For instance, I no longer include InbreedingCoefficient with my exome runs. This was spurred by a case where previously validated variants were getting discarded by VQSR. It turned out that these particular variants were homozygous alternate in the diseased samples and homozygous reference in the controls, yielding an InbreedingCoefficient very close to 1. We decided that the all-homozygous case was far more likely to be genuinely interesting than a sequencing/variant calling artifact, so we removed the annotation from VQSR. In order to catch the all-heterozygous case (which is more likely to be an error), we add a VariantFiltration pass for 'InbreedingCoefficient < -0.8' following ApplyRecalibration.
In my case, I think InbreedingCoefficient isn't as useful because my UG/VQSR cohorts tend to be smaller and less diverse than what the GATK team typically runs (and to be honest, I'm still not sure we're doing the best thing). Has anyone else found it useful to modify these annotations? It would be helpful if we could build a more complete picture of these metrics in a diverse set of experiments.