# AlphaMissense

Source: https://onco.cc/technologies/alphamissense/  
OnCo record `alphamissense` (Technology). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

Scored all 71 million possible single-letter protein changes in humans as likely harmful or benign.

## Summary

AlphaMissense is a Google DeepMind model derived from AlphaFold and fine-tuned on population variant frequencies to score whether a missense change is likely pathogenic or benign. The Science 2023 paper released classifications for all 71 million possible single-amino-acid substitutions in the human proteome, a complete catalogue rather than a tool that must be run per variant. Its scores are widely used to triage variants of uncertain significance in cancer genes, helping laboratories decide which findings deserve follow-up. The key caveat is that pathogenicity is not the same as actionability: a variant predicted to damage a protein does not tell a clinician whether a drug targets it or whether it changes management. For a newcomer: AlphaMissense has pre-scored every possible single-letter protein change so laboratories can see which ones look harmful.

## Fields

- Kind: Technology
- Status: established
- Last checked: 2026-09-08
- Tags: foundation-model; genome
- Principle: AlphaFold-derived model fine-tuned on population variant frequencies.
- Since: 2023
- Strengths: Complete catalogue
- Limitations: Pathogenicity is not the same as actionability

## Sources

- Science 2023: https://doi.org/10.1126/science.adg7492

## Connected records

- fronts: [AI & Computation](https://onco.cc/fronts/ai-computation/)
- companies: [Google DeepMind (and Google Research)](https://onco.cc/companies/google-deepmind/)
- terms: [Variant of uncertain significance (VUS)](https://onco.cc/terms/vus/)
- key papers: [Accurate proteome-wide missense variant effect prediction with AlphaMissense](https://onco.cc/key-papers/paper-cheng-science/)

---
JSON: https://onco.cc/api/v1/entities/alphamissense.json