Catalog of 71 million human missense mutations classified for clinical pathogenicity
AlphaMissense adapts AlphaFold to predict the clinical pathogenicity of all 71 million possible human single amino acid substitutions, categorizing 89% of variants with high confidence.
Comprehensive predictions for all 71 million possible human missense variantsTrained on population frequency data and structural context without clinical label leakageClassified 89% of missense variants as either likely pathogenic (32%) or likely benign (57%)Pre-computed database and downloadable BigWig/VCF files for integration into clinical pipelines
AlphaMissense adapts AlphaFold to predict the clinical pathogenicity of all 71 million possible human single amino acid substitutions, categorizing 89% of variants with high confidence.
Information checked against an official source; not a hands-on test. Source · Last reviewed: 20/09/2026, 11:09:40
Key Features
Comprehensive predictions for all 71 million possible human missense variants
Trained on population frequency data and structural context without clinical label leakage
Classified 89% of missense variants as either likely pathogenic (32%) or likely benign (57%)
Pre-computed database and downloadable BigWig/VCF files for integration into clinical pipelines
Academic Context & Research Evidence
Biological & Workflow Fit
Biological Application
Rare disease diagnosis, oncogenic driver mutation identification, and human genetics
Use this citation format when referencing AlphaMissense in scientific publications and benchmark papers.
@software{alphamissense_2026,
title = {{AlphaMissense}},
author = {{Google DeepMind}},
year = {2026},
url = {https://github.com/google-deepmind/alphamissense},
note = {Indexed on aibioatlas - AI for Biology and Drug Discovery}
}
Peer-Reviewed Literature & Preprints
Live scientific citations streamed from Europe PMC and PubMed for AlphaMissense.
⏳ Fetching real-time literature from Europe PMC & PubMed...
Technical / Product Information
Missing values mean the catalog has no recorded information. They do not mean a feature is absent.
Entry typeAI Model & Database
Access modeOpen Access
AI roleClinical Pathogenicity Classification
Input dataNot recorded
Output dataNot recorded
Licence conditionsCreative Commons Attribution 4.0 (CC BY 4.0)
Commercial eligibilityAvailable for research and non-commercial clinical evaluation
Compute requirementsPre-computed lookup table (database) / GPU for novel sequences
ValidationNot recorded
TypeStructural variant effect prediction model
Intended useNot recorded
CompatibilityNot recorded
ManufacturerGoogle DeepMind
Biological applicationRare disease diagnosis, oncogenic driver mutation identification, and human genetics
Research workflowQuery missense variant (e.g. BRCA1 p.Cys61Gly) -> obtain pathogenicity score (0 to 1)
Evidence levelPeer-reviewed research (Science 2023)