The Ensembl Variant Effect Predictor (VEP) is a powerful tool that allows users to input a list of genetic variants and determines which genes or regulatory elements they fall within, and what impact they may have. Ensembl VEP also reports the currently-known information about a variant location and scores from a wide range of variant deleteriousness predictors. This makes it very useful for helping to interpret whether a genetic variant is likely to be benign or deleterious, for example if it puts a person at risk of developing a certain rare genetic disease.

Google DeepMind has developed a new tool called AlphaMissense, which uses the AlphaFold human protein structure models to predict whether a sequence variant which changes one amino acid in a protein is likely to be tolerated or to impact protein function. 

The Ensembl Variant Effect Predictor now integrates Google DeepMind’s new AlphaMissense Catalogue. 

Variant pathogenicity scores based on protein structure have been developed in the past but these were hampered by the lower numbers of structures available. 

“Making such predictions easy to use in variant classification pipelines is a key part of what Ensembl VEP does so we have developed a plugin to integrate these scores, which can be used by anyone running their own installation of Ensembl VEP locally,” said Sarah Hunt, Variation Resources Coordinator at EMBL-EBI.

The database of predictions, which are freely available for non-commercial use, can be downloaded directly from the Google Cloud platform.  

“Clinical research studies and diagnostic labs are constantly discovering new variants when sequencing people with genetic diseases,” explained Andy Yates, Team Leader at EMBL-EBI. “We just don’t have enough experimental evidence to understand their impact, so we need robust and highly-accurate prediction tools. This is an exciting new tool which can help in the prioritisation of potential disease-associated variants.”

This integration is useful for a range of research applications, for example:

• Prioritising variants obtained from whole genome sequencing in a rare disease project
• Exploring the general landscape of what happens when you mutate proteins

Download AlphaMissense Catalogue.

Explore the Ensembl Variant Effect Predictor (VEP).

The AlphaMissense Catalogue is available only for the human genome references GRCh37 and GRCh38. Currently, there are no plans to expand it to other species. It is available under a CC-BY-NC-SA non-commercial licence. 

Edit