How mutect2 panel of normal works?
Answered
Dear GATK teams:
I am using GATK4 mutect2 with "PON" to call somatic variants in tumor-normal mode ,then I found some of the variants were disappeared in the raw VCF,some of the variants were targeted the label of "panel of normals" comparing with the same sample that didn't use "PON" .
a)What are the rules of targeting variants "panel of normal" or not ?
b)Why some variants in the raw VCF just be removed or added?
I have read a lot of material about "PON" ,but there is no detailed explanation on how a "PON" works in somatic calls. As far as I observed, I believe the targeting rules is not just present in the "PON" reference file.
Thank you,
Best wishes!
-
Hi, please note that your question was posted while the GATK Team was Out of Office.
Please repost any outstanding GATK issues and we will get to them if possible. Our first priority is solving GATK issues and abnormal results, see our support policy for more details.
-
I also observe similar differences in results obtained with and without PON. Two obtained lists of somatic variants are different, however, the overlap is high. I see additional variants in the list of somatic mutations with the use of PON (and those mutations were not observed at all when Mutect2 was run without PON).
I expected that the difference will be that I have pretty much the same list of somatic mutations but after using PON some of them will be tagged with the “panel_of_normals” flag but besides that, some mutations are completely lost and some new appear.
What is even more confusing some mutations that were PASS when analyzed without PON became flagged “strand_bias” after PON introduction.
Is Mutect2 run somehow differently if PON Is applied?
-
Hi Paulina GM,
We have some resources that go over how the PON works with Mutect2. You can also search more on this forum, I believe there have been similar discussions in the last year with some of our developers.
-
Hi Paulina GM, were any of those resources helpful and do you have any remaining questions that were not answered?
Thank you,
Genevieve
-
Paulina GM By default the PON blacklists variants before assembly as a speed optimization, and therefore variants in the PON usually don't show up in the output VCF. You can change this behavior with the -genotype-pon-sites argument. Under the default settings, some PON variants may still end up in the VCF if they are close enough to a non-PON variant to be assembled together.
As for the strand bias, FilterMutectCalls models the overall mutation rate of the tumor and its spectrum of variant allele fractions. Excluding the PON means that some artifacts will be included in this modeling, which in turn changes FilterMutectCalls' judgment about various filters.
-
Hi,
To follow-up on this question: does PON blacklist sites or specific variants? For example, if the PON has a A>C SNV, but the tumor has ALT reads supporting A>T, is this site labeled as PON and considered inactive? Or does it continue to reassembly and ignored by FilterMutectCalls PON filter?
This below from the FAQ suggests variants, but wanted to be sure because other language I have seen in documentation suggests sites.
"Mutect2 marks variants that are found in the PON with the “PON” info field, which FilterMutectCalls then uses for filtering. Additionally, Mutect2 considers variants in the PON as inactive by default (this can be changed with the -genotype-pon-sites argument), so most will be silently pre-filtered without ending up in the output. Some PON sites are output because they may appear in an active region containing a non-PON site."
-
TA The PON blacklists sites. Any variant allele at a PON site is rejected. This is intentional (although it is not always correct and can lead to false negatives) because often sites do have a general tendency toward error and not just a particular allele. This is especially true in the case of mapping error.
-
Got it thank you for your quick response!
Please sign in to leave a comment.
8 comments