How to differentiate between protective and damaging variants

Question

[praj_bonjour]

Hello,

My project involves to detect variants than influences lifespan but I am not sure how to differentiate between protective and damaging variants. I am analysing rare variants, for which I have also used SIFT and Polyphen. The tools predict the variants as 'damaging' or 'benign'. As far as I understand 'damaging' may not necessary be a negative impact,it just shows that there is a change in function- right ?

For a bioinformatician, what analysis needs to be performed to know if the variants are damaging or protective ?

Thank you.

Answer 1

[nilakshafreezon]

1. I'm not a very experienced bioinformatician but as per my knowledge 'damaging' in Polyphene and SIFT does not simply implicit that it has a change in functional protein, that are in turn called nonsynonymous SNPs. Both polyphene and SIFT uses special algorithms to predict the probability of potential damage caused by a variant, and gives a score and using thresholds those scores are either fallen to Benign or Damaging. You may get all these plus more information from their websites. Try PROVEAN as well.
2. I guess you might need to query against clincal databases such as ClinVar to annotate your variants against the clinical phenotype which in turns would tell you whether your variants are pathogenic, and if so to what extent using publicly available data submitted using databases like OMIM and MedGen.

Folks please correct me if I was wrong at any point :)

Regards.

Answer 2

[DG]

Firstly, always take the predictions of PolyPhen, SIFT, etc with a grain of salt. In general they are attempting to predict deleterious functional impacts on a protein but in theory, some of these mutations could actually be beneficial. However, this is only likely to be a very very tiny fraction of all such sites. In general most of the mutations predicted to be damaging likely have little to no effect on protein function and some of those predicted to be benign will actually have a functional impact. There are a few papers out there comparing the predictions of various tools to each other to show how much they agree/disagree with one another.

In general much of the work out there, is analyzing variants in the context of negative effects. But there are ones out there looking for positive effects. I believe there have already been GWAS studies done on longevity for instance and possibly even some exome/genome studies. It is hard to give specific advice other than "read everything you can for any studies even remotely similar in study design" without knowing how you are approaching the problem already (panels, exome, genome, etc)