AlphaMissense

Overview

AlphaMissense is a deep-learning pathogenicity predictor for amino acid substitutions developed by Google DeepMind. It leverages the AlphaFold2 protein structure framework combined with evolutionary and population-genetic features to classify missense variants as likely benign, ambiguous, or likely pathogenic. It achieves high discrimination between disease-causing and neutral variants across a wide range of genes.

Used by

• Applied to score functional impact of somatic variants in endometrial polyps; UBE2A p.(Arg6Trp) received an AlphaMissense score of 0.96 (likely pathogenic), supporting its nomination as a driver alteration PMID:28445112

Notes

• Scores range from 0 (benign) to 1 (pathogenic); threshold of ~0.56 is commonly used for “likely pathogenic.”
• Best used in conjunction with other evidence (CADD, REVEL, population frequency, functional data).
• Pre-computed scores available for all possible human missense variants.

Sources

• PMID:28445112

This page was processed by crosslinker on 2026-05-14.