Rating variants for their potential deleteriousness is vital for solving the link between genotypes and phenotypes. There are many different algorithms for predicting how likely it is that a human variant would affect the function of a protein, and in release 94 of Ensembl, we’ll be making more of these available.

Currently, we have SIFT and PolyPhen-2 results available for human missense variants and for our next release, we’re also bringing in CADD, REVEL, MetaLR and MutationAssessor. These will be visible in the gene and transcript variant tables, and in the genes table for a variant.

We are using our traffic light system to give a quick overview of how variants are rated by the different tools but recommend the individual scores are used for detailed variant analysis. The tables on our gene and transcript pages make it easy to pick your own cut-off for your prefered algorithm.

A word of warning

It’s important to remember that these predictors of pathogenicity are just that, predictors. They are not evidence of pathogenicity and should only be used as a guide for further experiments or analysis. Because they all work in slightly different ways, you will get different predictions from each one.