VQSRTranche filters in GATK

Question

[anu014]

Greeting of the day Biostars!

I'm dealing with variant calling in WES data nowadays. For this, I'm using GATK ,most trusted tool for variant calling. But how to know which is the true variant & which one is false positive? For this I learned there's something called 'VQSRTranche' filters & it goes like this :

FILTER=<ID=VQSRTrancheSNP99.00to99.90,Description="Truth sensitivity tranche level for SNP model at VQS Lod: -1.3154 <= x < 0.5827">

So following are the questions I've on my mind :

i. What is tranche level/tranche?
ii. Eg. VQSRTrancheSNP99.00to99.90 means the variant falls in confidence level 99.00 to 99.9 WITH false positive % being 0.1?
iii. 'VQS Lod' value for same trance changes from data to data.
iv. '--ts_filter_level 99' marks tranche level below 99 as PASS/TRUE variant?

[Pierre Lindenbaum]

read the doc: https://gatkforums.broadinstitute.org/gatk/discussion/39/variant-quality-score-recalibration-vqsr

Answer 1

[Raony Guimarães]

This image explains more about the tranches

[Pierre Lindenbaum]

This image explains more about the tranches

IMHO it doesn't explain anything. A legend is missing...

Answer 2

[Nawar]

I hope you find this post useful:

Variant Quality Score Recalibration (VQSR)