Question regarding version 2.4 PureCN

[maxanes]

Hi,

We are using version bioconductor-purecn V 2.0.2 for our analysis of the selection of clonal mutations, recently we have tried V 2.4.0 and noticed that many mutations are filtered out specifically ones that have BQ < 25, but our VCF file has only MBQ, meaning that uses that for filtering?
The problem is that those are mutations relevant to us such as KRAS/TP53 and so on. We have tried to set min.supporting reds =0 but didn't turn off this filter. Have you any comments/recommendations for the version to use?
Attached are log files for the same sample run with both versions
C14_ffpe.purecn_V2.0.2.log
C14_ffpe.purecn_V2.4.0.log
Thank you in advance.

[lima1]

I can add an option to change the 25 today. The default is pretty low though, so you might want to try to remove the artifacts among them upstream.

[maxanes]

I am not sure about this threshold of 20 or 25, does that mean that in the previous version 2.02 variants such as KRAS mutation called clonal with MBQ 20 are not true?
It seems that V2.02 ends up with 19340 variants to use while V2.4 uses only 5002 variants.

[lima1]

5000 variants is wrong for a WES panel, should be at least 20000 heterozygous SNPs. Are you using Mutect2 in a recent GATK4? Any steps that deviate from their best practices? Do you use the baits file (location of baits, not exons) in IntervalFile.R? Do you run Mutect with interval padding to get SNPs in introns? With your coverage, you can use 100bp padding.

[maxanes]

I also think that 5000 is too little, yes we use Mutect2 in a recent GATK4, the same baits file (not sure which of those, might be exons) I use for both versions and running it in the same way, Mutect with interval padding 50

[lima1]

Any deviation from here: https://gatk.broadinstitute.org/hc/en-us/articles/360035531132--How-to-Call-somatic-mutations-using-GATK4-Mutect2 ?

The contamination step is not critical, but in general worth it. Same with the baits file. The baits locations give you the cleanest signal.

[maxanes]

The only thing that is different is that we use a normal sample instead of pon for Mutect2, and the DB annotation that we do afterward.

[lima1]

The PoN is a powerful way of removing artifacts, I would consider it. Your input VCF contains a lot of artifacts. I can't remember if Mutect2 is using GnomAD in the likelihood model, it might.

[lima1]

And not forget using the --genotype-germline-sites with matched normals (I'm sure you do).

[lima1]

Also likely hit by #320.

[maxanes]

Thanks 😊 From: Markus Riester ***@***.***> Sent: Saturday, September 9, 2023 4:24 PM To: lima1/PureCN ***@***.***> Cc: Marijana Nesic ***@***.***>; Author ***@***.***> Subject: Re: [lima1/PureCN] Question regarding version 2.4 PureCN (Issue #311) Also likely hit by #320<#320>. — Reply to this email directly, view it on GitHub<#311 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AV6BKZNP2GPCFEG4Z5L3Y7DXZR3X7ANCNFSM6AAAAAA2YOF72I>. You are receiving this because you authored the thread.Message ID: ***@***.******@***.***>>