VariantRecalibrator latest recommandations
Answered
Hello,
I am currently trying to perform germline variant calling for human, however, I am somewhat stuck at the VQSR step. A big problem is, that all the links ("method documentation", "parameter recommendations" and "tutorial") for the current version of the VaraintRecalibrator are dead, i.e. just lead to the homepage. Thus, I dont really know what to do, and what options I should use, e.g. for the prior (I barley find any documentation what exactly that is, or what a good default value is).
Furthermore, I am having difficulties, because I am using the GRCh37 reference genome (hs37d5.fa) and I can only find "high confidence" SNPs for the GRCh38 reference genome in the resource bundle from the Broad.
Is it possible to simply download all 1000g variants (from here), combine and filter them for SNP and INDELs respectively, and then use them as the training set?
I would be very grateful is somebody could help me out.
Cheers
-
Hi nhaus,
I am going to move your post into our Community Discussions -> Documentation Questions topic, as the Germline topic is for reporting bugs and issues with GATK.
You can read more about our forum guidelines and the topics here: Forum Guidelines.
Best,
Genevieve
-
Hi nhaus,
Thanks for bringing these broken links to our attention, we have been trying to get everything sorted out since we moved our website and some things are still not working!
Here is what you were looking for:
- The method documentation: Variant Quality Score Recalibration (VQSR)
- Tutorial: (How to) Filter variants either with VQSR or by hard-filtering
- I don't think we still have any document regarding the parameter recommendations but some recommendations can be found in those two articles above
If you still have more questions regarding VQSR after reading those two articles, please follow up. I think they may have recommendations regarding the training set.
Best,
Genevieve
-
nhaus I created a PR to fix these docs: https://github.com/broadinstitute/gatk/pull/7270. Thank you!
Please sign in to leave a comment.
3 comments