Hi everyone. Is there any pipeline for prioritisin...
# variantprioritization
m
Hi everyone. Is there any pipeline for prioritising germline variants, maybe including the ACMG classifications? Are you aware of any tools/pipelines for such a task?
I am aware of #raredisease using GENMOD but it seems mainly suitable for family trio / pedigree analyses and not single-sample
l
VEP is a pretty versatile tool that is running in several of the nf-core pipelines, including Sarek. Check out the customizations of it, it can all be done through nextflow config-files piping into the VEP world. And that's big
m
VEP is just a variant annotation tool, surely with lots of resources, but it doesn't prioritize variants according to some well-thought guidelines, such as ACMG. Sorry if my question wasn't clear enough!
l
I know. I also know that ACMG is too fuzzy to actually algorithmify. There's some off-hand attempts and papers out there, but nothing serious. So VEP is the closest in my opinion.
m
I think there is nothing serious out there because most of the valid algorithms are IP protected and have been turned into companies or services. ML/AI has done a lot in the space but I was wondering if there is any open source tool that would be a good starting point.
l
Mmmmh... that's your opinion. Maybe. Good luck any way! 😄
And really - VEP is amazing, check it out
m
Well, sure, just my two cents! I'm not saying I have the absolute truth. I know VEP is amazing, but it doesn't cover some aspects of interest in variant prioritisation
l
My cents too, agreed. And it's an interesting discussion. Thanks. Would love to hear what other's in the channel think too.
a
Hi Matteo. Ive been working similar problems. Right now, I’m using sarek to obtain ClinVar/VEP annotated VCFs and then querying the variants either manually or via an API call. Variant interpretation and prioritisation according to AMP/ACMG guidelines is always cumbersome and requires some level of manual review. I am the using a combination of open source databases (my applications are in cancer genomics) 1. CIViC 2. OncoKB 3. Cancer Genome Interpreter 4. MTB Portal (karolinska) There is also VarSome which is a commercial offering. Unfortunately, these databases are sometimes hard to query and may provide conflicting results as they are maintained by independent groups/ consortia. I’m not sure if it is feasible to intergrate automatic database annotations into the nfcore sarek pipeline, but if possible it would surely be helpful
This paper has a good comparison of the various knowledge bases for somatic variant interpretation https://www.nature.com/articles/s41588-020-0603-8 Part of my work is analysing germline variants in lung cancer patients which predispose them to cancer. However, I’m still not aware of much for germline besides ClinVar
👍 1
m
Hi Abhay! I'll have a look into these resources. Thanks for contributing!
r
I would actually recommend the cancer genome interpreter, their algorithm is quite good and the API works well. In their list of mutations they separate between germline and somatic which I think might be useful for you. This would be the most practical and direct solution because as the paper says there is a need for standardised curation but yeah, not an easy solution for that one. I haven’t used the proposed harmonise solution in the paper but I have used cgi for years without any issues
👍 3
m
Thanks Raquel!