What is the correct step? VQSR and CNNScoreVariants?

Answered

Follow

Damian Fermin

• September 04, 2020 17:29

I think this should be an easy one so I apologize if it's a stupid question.

I'm confused from reading the Best practices workflow for Germline short variant discovery.

I am trying to get a set of "analysis-ready variants".

I've run ApplyVQSR on my VCF file. What step comes next? Am I suppose to now run CNNScoreVariants on my VCF file? Or do I run FilterVariantTranches?

I can't tell if that is now the assumed best practice or if I should just run CNNScoreVariants only. 

Thanks in advance for any and all help. 

0

• 

  Kshama Aswath

  • September 05, 2020 01:08

  Hi Damian,

  I am a pretty new user of GATK pipeline myself and am drowned in reading documentation after documentation and many times end up finding out "new steps" to use accidentally on GATK search or google search or through blogs etc.

  There is tremendous amount of information out there and we always have to choose what best fits our final goal.

  Having said that, I can share what  I have figured out.

  Hopefully sharing this helps us both :)

  For germline analysis, on a large cohort > 100 exomes, that I have, I did VQSR (assuming you did everything in order until this according to the GATK best practices) then "apply VQSR". You could use Calculate genotype posteriors  ( you must have genotype likelihood info from haplotype caller to do this) to sort of further assess and quantify uncertain genotype calls , then redo variant filtration and variant annotator  to get the ones that are not called by GATK/ or not in known spectrum aka novel ones out and then take it through Michigan imputation to validate it further.

  I found this paper super useful in understanding more, check this out: 

  https://core.ac.uk/download/pdf/323064116.pdf

  This GATk document also helps with further steps:

  https://gatk.broadinstitute.org/hc/en-us/articles/360035890551-Allele-specific-annotation-and-filtering-of-germline-short-variants

  0

  Comment actions Permalink
• 

  Bhanu Gandham

  • September 09, 2020 18:35
  • Edited

  Kshama Aswath

   

  Thank you so much for sharing your GATK experience here. Your interactions and suggestions will help build the knowledge base of this forum and help other scientists with their research :)

  The GATK team greatly appreciates your contribution!

  0

  Comment actions Permalink

Please sign in to leave a comment.