Question
Question Posted 06/04/20:
Hi, for ADNI WGS VCFs, it has been mentioned that GATK VQSR (VariantRecalibrator + ApplyVQSR) variant QC steps were used. Can you please tell me how these were applied? For instance, which "--truth-sensitivity-filter-level" was used for SNPs and Indels? I could not find this information.
Thank you for your attention.
Hi, for ADNI WGS VCFs, it has been mentioned that GATK VQSR (VariantRecalibrator + ApplyVQSR) variant QC steps were used. Can you please tell me how these were applied? For instance, which "--truth-sensitivity-filter-level" was used for SNPs and Indels? I could not find this information.
Thank you for your attention.
Response posted 06/04/20 by Kwangsik Nho:
In the ADNI WGS analysis, the best Practices for Variant Calling with the GATK was applied as default parameters.
Response posted 06/05/20 by Fahri Küçükali:
Thanks! Indeed, that is indicated in the PDF. However, as far as I can see GATK v3.1 was used at that time. It is trivial to access the Best Practices for GATK v3.1 as it is not supported anymore, but nevertheless the Best Practices usually does not suggest a specific "--truth-sensitivity-filter-level" to be used for the ApplyVQSR step after recalibration
The current tool specifications uses 99.0 for SNP truth sensitivity filter level (https://gatk.broadinstitute.org/hc/en-us/articles/360042477792-ApplyVQSR) as an example. I vaguely remember that 99.5 was commonly used in the times of GATK v3.
But anyway, looking at the header of ADNI WGS VCFs (GATK-processed), I can see that recalibration was made for only 99.7, 99.8, 99.9, 100.0 levels. Therefore one of these 4 levels should have been used, but I am not sure which one. This is important, and I wonder if there is a way to know about this?
Thanks again.
The current tool specifications uses 99.0 for SNP truth sensitivity filter level (https://gatk.broadinstitute.org/hc/en-us/articles/360042477792-ApplyVQSR) as an example. I vaguely remember that 99.5 was commonly used in the times of GATK v3.
But anyway, looking at the header of ADNI WGS VCFs (GATK-processed), I can see that recalibration was made for only 99.7, 99.8, 99.9, 100.0 levels. Therefore one of these 4 levels should have been used, but I am not sure which one. This is important, and I wonder if there is a way to know about this?
Thanks again.
Response posted 06/05/20 by Kwangsik Nho:
It seems the ADNI WGS analysis pipeline used 99.7 for both SNP and INDEL.
