search

sigven / cpsr

Cancer Predisposition Sequencing Reporter (CPSR)
https://sigven.github.io/cpsr/
Other
56 stars 12 forks source link

ClinVar conflicting evidence #9

Closed vladsavelyev closed 5 years ago

vladsavelyev commented 5 years ago

Hi Sigve,

Wondering what was the rationale behind skipping variants which have "conflicting_interpretations_of_pathogenicity" by ClinVar regardless of other annotations that can be (likely) pathogenic? Perhaps it makes the report too cluttered?

Thinking that such variants should still go under TIER3 (either VUS or Non-classified) - or below, but should show up in the report.

Perhaps in the following line, the check is.na(CLINVAR_CLINICAL_SIGNIFICANCE) should be removed?

https://github.com/sigven/pcgr/blob/b6ccae7fac4c86c472fdac6e0c9728edd930a189/src/R/pcgrr/R/cpsr.R#L169

cc @ohofmann

vladsavelyev commented 5 years ago

Attaching an example VCF with such variant: CHEK2.vcf.gz

sigven commented 5 years ago

Hi Vlad, Check out the new release, should place variants with conflicting interpretations of pathogenicity (from ClinVar) in the tier 3 table.

cheers, Sigve

vladsavelyev commented 5 years ago

Wow, thanks so much for the release, Sigve! That's highly anticipated. Will give it it run

sigven commented 5 years ago

Cool. Happy bug-hunting :D

vladsavelyev commented 5 years ago

Had to adapt for the command line options, but otherwise no bugs noticed :)

One thing is that you removed the predisposition gene list from the toml. Currently we pre-subset the germline variants to our own gene list (which is a combination of your list and CancerGeneCensus germline) before feeding them into CPSR, so in a way wanted to avoid additional hard-filtering. Though that's not critical I think given that overall improvements are so significant.

Also check out the updates to PCGR as well. Nice that it reports purity and ploidy, I will take them from PURPLE caller that we use for CN calling. But curious if purity/ploidy are used only to show in the report, or also for filtering somehow?

Also you made tumor_type required now, but we don't yet specify it in our pipeline on a regular basis. I will just set it to Cancer_Unknown_Primary_NOS, do you think it is fine for a general case? A lot of our tumors are CUPs anyway and that's why we don't even parameterize our analysis by tumor type.

vladsavelyev commented 5 years ago

And thanks for parametrizing vep_pick_order :)

sigven commented 5 years ago

Yeah, the list of genes that are target for predisposition screening was simply a choice I made. The previous configuration was a bit too messy to handle. Either way, I'd be happy to create the specific gene set you are using as an additional virtual panel (aka UMCCR track:-))

Regarding ploidy and purity you are correct; these are not used in any way other than for display right now (I have no big clue as to how they could be used for filtering, as the purity and ploidy preferably should be taken into account when considering the variant set that is fed as input.

Wrt tumor type I believe I should add the option of not specifying any type, the 'Cancer_Unknown_Primary_NOS' is in fact a distinct cancer subtype, and should not be confused with not having specified a type at all. Sorry for not considering this.

Best, Sigve

vladsavelyev commented 5 years ago

Cool. Would be really awesome to have the option of generic cancer type!