Integration of AlphaMissense scores into Ensembl

by ·Comments Off on Integration of AlphaMissense scores into Ensembl

We’re excited to announce the integration of AlphaMissense pathogenicity scores into Ensembl! The display of AlphaMissense data in Ensembl and other EMBL-EBI resources aims to empower scientists to gain new insights, expand the exploration of genetic variation and the tolerance to change of different regions in proteins.

AlphaMissense is an AI model developed by Google DeepMind which classifies genetic variants, specifically missense variants, as more likely to be pathogenic or benign. This information is helpful when researching variant-disease associations and provides an indication of the most functionally important parts of a protein.

AlphaMissense scores are now integrated into the Ensembl Variant Effect Predictor (VEP) tool, enabling easy annotation of variants via the web interface, REST API, or command-line interface. Average AlphaMissense pathogenicity scores for each amino acid can also be visualised on the AlphaFold predicted 3D protein structure, available from the associated Ensembl transcript page.

Ensembl VEP

You can enable AlphaMissense scores on Ensembl VEP as follows:

Web interface

Open Ensembl VEP on the browser, scroll down to ‘Additional configurations’ and expand the ‘Predictions’ tab where you can enable AlphaMissense scores.

A screenshot of the Ensembl VEP web interface input form highlighting how to retrieve AlphaMissense scores. Under 'Additional configurations', open the 'Predictions' tab and select 'AlphaMissense'.
A screenshot of how to enable AlphaMissense scores in the Ensembl VEP web interface input form.

The AlphaMissense pathogenicity score and classification will be reported in separate fields in the Ensembl VEP output table:

A screenshot of an example output table of an Ensembl VEP web interface query. AlphaMissense classification and scores can be found in separate columns in the output table.
A screenshot of an example output table including ‘AlphaMissense classification’ and ‘AlphaMissense pathogenicity score’ in separate columns.

REST API

AlphaMissense scores and pathogenicity classifications can be enabled by adding the optional parameter AlphaMissense=1 when querying any /vep/ endpoint.

Command-line

On the command-line Ensembl VEP, you can use the AlphaMissense plug-in to analyse your data locally. Read more about how to use plug-ins on the Ensembl VEP documentation page.

AlphaMissense scores in the Ensembl browser

Average AlphaMissense pathogenicity scores for each amino acid can also be visualised on the AlphaFold predicted 3D protein structure, available under ‘Transcript-based displays: Protein Information’ and selecting the ‘AlphaFold predicted model’ display.

A screenshot of the 'Transcript' tab for the human CLINT1 gene. AlphaMissense scores can be visualised by opening the 'Protein Information: AlphaFold predicted model' display.
A screenshot of the ‘Transcript’ tab of the human CLINT1-201 transcript highlighting how to visualise AlphaMissense scores in the AlphaFold predicted protein structure.

This interactive view allows users to switch between variants, domains, exons, and AlphaMissense results, supporting the interpretation of different regions of the protein and their sensitivity to change.

A screenshot of the 'AlphaFold predicted model' for the human CLINT1-201 transcript. AlphaMissense scores can be visualised in the protein model by selecting 'AlphaMissense Pathogenicity' under 'Toggle' in the right-hand panel. Likely pathogenic regions are highlighted in red in the protein structure and likely benign regions are highlighted in blue.
A screenshot of the ‘AlphaFold predicted model’ for the human CLINT1-201 transcript, highlighting where to enable AlphaMissense scores. Likely pathogenic regions are highlighted in red in the protein structure and likely benign regions are highlighted in blue.

AlphaMissense data are also available in other EMBL-EBI resources UniProt and the AlphaFold Protein Structure Database. You can read more about AlphaMissense in the 2023 publication by Cheng, et al. and its availability in EMBL-EBI resources in the blog post ‘AlphaMissense data integrated into Ensembl, UniProt and AlphaFold DB’.

Author: Louisse Paola Mirabueno
Editors: Sarah Hunt, Benjamin Moore, Roz Onions, Jamie Allen