CombineVariants --removeFILTER feature request

tommycarstensentommycarstensen United KingdomPosts: 152Member

When using CombineVariants, my variants get a FILTER value of either PASS or LowQual. Would it be possible to add an option to CombineVariants which prevents the FILTER value to be set to PASS? Otherwise I have to do some file processing before I run ApplyRecalibration further downstream. It would be great if this was a feature of all walkers and not just VariantFiltration. I'm not sure if the forum is the right place for feature requests. Happy to use Bugzilla or similar instead. Thanks.

Best Answer

Answers

  • ebanksebanks Posts: 684GATK Developer mod

    Hi Tommy,

    I'm a bit confused - why do you need to do processing before Apply Recalibration to handle PASSing filter fields? Are you saying that it's incorrectly turning filtered records ("LowQual") into PASSing ones?

    Eric Banks, PhD -- Senior Group Leader, MPG Analysis, Broad Institute of Harvard and MIT

  • tommycarstensentommycarstensen United KingdomPosts: 152Member

    Hi Eric,

    LowQual sites are not turned into PASSed sites; at least I haven't seen it for the few entries I checked.

    Here is what happens. CombineVariants writes the FILTER value PASS to SNPs and INDELs. ApplyRecalibration will not modify/update the FILTER value of INDELs when VQSR is run on SNPs and vice versa. It's confusing if you look at the AR VCF output. I couldn't understand, why my INDELs were PASSed. It's not a big problem in my case though, because ProduceBeagleInput follows AR in my case and PBI only emits SNPs anyway. In the meantime I ended up writing my own tool to replace the functionalities of ApplyRecalibration and ProduceBeagleInput to save looping over the same data twice and saving the same data to disk twice.

    Thanks and sorry for posting my request. If I had known it was only INDELs left as PASSed, then I wouldn't have asked for the feature. Thanks.

  • ebanksebanks Posts: 684GATK Developer mod

    Sorry for being thick here, but I still don't actually understand the problem. What is it exactly that isn't working with PASS in the FILTER field? Thanks.

    Eric Banks, PhD -- Senior Group Leader, MPG Analysis, Broad Institute of Harvard and MIT

  • tommycarstensentommycarstensen United KingdomPosts: 152Member

    Hi Eric,

    I didn't explain very well.

    After CombineVariants I have this INDEL: 1 10106 . C CA 682.56 PASS

    After ApplyRecalibration --mode SNP I have this INDEL: 1 10106 . C CA 682.56 PASS

    The INDEL obviously didn't PASS the VQSR filtering step run on SNPs, but rather the FILTER field is a remnant of running CombineVariants. It just confused me, when I saw it. It's not a problem for my downstream analysis.

    Thanks.

  • ebanksebanks Posts: 684GATK Developer mod
    Answer ✓

    Ah, now I understand. Okay, I'll log a report for this. It's not a high priority item obviously but we'll try to fix it before the next release. Thanks!

    Eric Banks, PhD -- Senior Group Leader, MPG Analysis, Broad Institute of Harvard and MIT

Sign In or Register to comment.