variant calls which are important

Question

[kanwarjag]

I used a VCF file to filter somatic mutations based on various various filters as suggested by vendor such as Pass/ Fail based on somatic score, quality score, read depth etc. I ma down to around 50 variants which has various different functions and are map to different genomic locations - introns, exons, promoter etc. some of them are predicted to be associated with loss of function while other will not affect any such change. The question, I am struggling with is which variants may be important to report or validate. Are only exonic variants important or I should report the once which are associated with loss of function. The data is comparing matched normal vs tumor samples with underlying aim to identify the disease associated genes. Any suggestion or pointers, links to study further will be helpful.

Thanks

[kanwarjag]

Thanks Devon, I agree that one has to move to downstream analysis. I am not sure how 3' UTR variation may be predicted to be associated with loss of function.

Answer 1

[Devon Ryan]

Exonic mutations are generally more interesting, since it's often simpler to predict what they might do. Having said that, a variant in a gene with no plausible affiliation to cancer (or whatever else you happen to work on) is probably not worth following up on. At some point you have to move from bioinformatic predictions to understanding the underlying biology to progress in prioritizing mutations. So it depends on the goals. If your goal is to create a ranked list of interesting changes to look at then prioritize things by how deleterious it's predicted to be and how plausibly linked the affected gene is to the disease.

[DG]

I'll add that since you are looking at cancer there are a bunch of options available for doing this prioritization. Ultimately you will be reporting the most well supported/interesting candidates, but should also probably report all 50 of those filtered variants in some sort of supplementary material. There may be something interesting there you missed that will be followed up on by other groups after all.

Some of the tools/sites to use for prioritizing include Cosmic (report any known cancer mutations or new mutations in known cancer-associated genes). For predicting functional impacts use several tools, not just one. PolyPhen, EvoD, FATHMM, etc.

Intronic or UTR variants may be doing something but they tend to be much harder to predict. You can do some searching to see if they lie within known miRNA binding sites, which is somewhat reasonable with only a handful of variants.

Answer 2

[swbarnes2]

What you are asking is pretty much the question of this decade of biology...now that we can find SNVs easily, how do we figure out what they mean? So there is no easy answer. Some of it will need to be answered by someone who actually knows a lot about the biology of what you are studying, which may or may not be yourself. Depending on how long your list is, you might want to focus on deleterious amino acid changes, or all amino acid changes, mutations in splice sites, and any kind of mutation in a gene that the literature suggests is involved in whatever you are studying. Some groups will not want to bother investigating mutations deep in introns (unless they might make a new splice site), or in putative promoters, or in 3' UTR.

Answer 3

[datanerd]

Not exactly an answer to the specific question, but this is paper is very interesting and I think a good one for everyone working to find disease associated gene/variants.

http://www.nature.com/nature/journal/v508/n7497/full/nature13127.html