SEQanswers

SEQanswers (http://seqanswers.com/forums/index.php)
-   Bioinformatics (http://seqanswers.com/forums/forumdisplay.php?f=18)
-   -   Gatk vqsr (http://seqanswers.com/forums/showthread.php?t=40914)

variadevang 02-14-2014 04:11 AM

Gatk vqsr
 
What does VQSR actually do? It improves the score? It filters the variants? or exactly what?
Is it an alternative for VariantFiltering based on hard values? i desperately need help on this.
p.s. i have gone through pdf's, tutorials, the complete GATK website but have failed to understand the concept.

IF POSSIBLE PLEASE EXPLAIN IN A BIT BRIEF

grakocevic 02-15-2014 02:48 PM

It filters the variants.

Let's start from the plain old "hard" filters as you're probably used to.
We have a bunch of metrics for each variant, e.g. call quality, coverage depth at the position, mapping qualities (average across reads or their rank sum), etc. So with filtering what we do is set some thresholds and filter out anything that does not meet those thresholds.

The problem with that is figuring out those thresholds. Broad has some recommendations, but those are mostly based on the average of data they've seen. For your particular dataset you may need to set some of those higher or lower.

VQSR tries to that for you. It takes a bunch of variant calls that it assumes to be true (e.g. calls that match know variation sites through HapMap), looks at the different metrics and the distribution they take, and builds a model of what a true positive variant looks like.
(this would be an approximate overview of the idea…)

It is an alternative to VariantFiltering based on hard values. However, you will need a certain amount of data to make it usable. WGS are fine, whole exomes work (but can sometimes be a pain to get the params right), anything smaller you'd probably need to do hard filters.

variadevang 02-19-2014 01:33 AM

grakocevic
Thanks a lot for the explanation. But i did not understand that if it filters, then why are the number of variant calls same as before recalibration?

grakocevic 02-19-2014 07:25 AM

You should check the Filter column in the VCF.
Filtering variants in a VCF usually doesn't mean deleting them, but rather writing something in that column.
Anything that is not '.' or 'PASS' typically means that the variant is filtered, and many tools will actually disregard all such variants.

If really you want to remove the filtered variants you can use GATK SelectVaraints and set it to exclude filtered...


All times are GMT -8. The time now is 10:40 PM.

Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2020, vBulletin Solutions, Inc.