Published September 19, 2023 | Version v3
Journal article Open

Predictions for AlphaMissense


This repository provide AlphaMissense predictions. For questions about AlphaMissense or the prediction Database please email 

File descriptions

AlphaMissense_hg19.tsv.gz, AlphaMissense_hg38.tsv.gz

Predictions for all possible single nucleotide missense variants (71M) from 19k human protein-coding genes (canonical transcripts) for both hg19 and hg38 coordinates. These files are sorted by genomic coordinates.

AlphaMissense_gene_hg19.tsv.gz, AlphaMissense_gene_hg38.tsv.gz

Gene-level average predictions, which were computed by taking the mean alphamissense_pathogenicity over all possible missense variants in a transcript (canonical transcript). 


Predictions for all possible single amino acid substitutions within 20k UniProt canonical isoforms (216M protein variants). These are a superset of the amino acid substitutions induced by single nucleotide missense variants. This file uses UniProt accession numbers for proteins and does not have genomic coordinates. 


Predictions for all possible missense variants for 60k non-canonical transcript isoforms (hg38, GENCODE V32). This file has transcript_id but no UniProt accession numbers. Predictions for non-canonical isoforms were not thoroughly evaluated and should be used with caution. This file is sorted by genomic coordinates.


Predictions for all possible single amino acid substitutions for 60k non-canonical transcript isoforms (GENCODE V32). These are a superset of the amino acid substitutions induced by single nucleotide missense variants.This file has transcript_id but no UniProt accession numbers. 

All transcript annotations are based on GENCODE V27 (hg19) or V32 (hg38). Canonical transcripts are defined as described in the publication.

All files are compressed with bgzip.

Column descriptions  

Note that not all columns are present in every file.


The chromosome as a string: chr<N>, where N is one of [1-22, X, Y, M].


Genome position (1-based).


The reference nucleotide (GRCh38.p13 for hg38, GRCh37.p13 for hg19).


The alternative nucleotide.


The genome build, hg38 or hg19.


UniProtKB accession number of the protein in which the variant induces a single amino-acid substitution (UniProt release 2021_02).


Ensembl transcript ID from GENCODE V27 (hg19) or V32 (hg38).


Amino acid change induced by the alternative allele, in the format <Reference amino acid><POS_aa><Alternative amino acid> (e.g. V2L). POS_aa is the 1-based position of the residue within the protein amino acid sequence.


Calibrated AlphaMissense pathogenicity scores (ranging between 0 and 1), which can be interpreted as the predicted probability of a variant being clinically pathogenic.


Classification of the protein_variant into one of three discrete categories: 'likely_benign', 'likely_pathogenic', or 'ambiguous'. These are derived using the following thresholds: 'likely_benign' if alphamissense_pathogenicity < 0.34; 'likely_pathogenic' if alphamissense_pathogenicity > 0.564; and 'ambiguous' otherwise.


The average alphamissense_pathogenicity of all missense variants per transcript.


Copyright (2023) DeepMind Technologies Limited

All materials are licensed under the Creative Commons Attribution 4.0 International License (CC-BY) (the “License”).  You may obtain a copy of the License at:

Unless required by applicable law or agreed to in writing, all materials distributed under the License are distributed on an "AS IS" AND “AS AVAILABLE” BASIS, WITHOUT REPRESENTATIONS, WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Researchers interested in predictions not yet provided can send an expression of interest to 


This is not an officially supported Google product

The AlphaMissense Database contains predictions with varying levels of confidence, caution should be exercised in use. The information provided is not intended to be a substitute for professional medical advice, diagnosis, or treatment, and does not constitute medical or other professional advice. AlphaMissense has not been validated for, and is not approved for, any clinical use.


If you use this resource for your research please cite the following publication:

“Accurate proteome-wide missense variant effect prediction with AlphaMissense”

Jun Cheng, Guido Novati, Joshua Pan, Clare Bycroft, Akvilė Žemgulytė, Taylor Applebaum, Alexander Pritzel, Lai Hong Wong, Michal Zielinski, Tobias Sargeant, Rosalia G. Schneider, Andrew W. Senior, John Jumper, Demis Hassabis, Pushmeet Kohli, Žiga Avsec 

Data format samples

AlphaMissense_hg19.tsv.gz, AlphaMissense_hg38.tsv.gz

#CHROM  POS        REF    ALT       genome        uniprot_id  transcript_id      protein_variant  am_pathogenicity  am_class

chr1    69094      G      T         hg38          Q8NH21      ENST00000335137.4  V2L              0.2937            likely_benign

chr1    69094      G      C         hg38          Q8NH21      ENST00000335137.4  V2L              0.2937            likely_benign

chr1    69094      G      A         hg38          Q8NH21      ENST00000335137.4  V2M              0.3296            likely_benign

chr1    69095      T      C         hg38          Q8NH21      ENST00000335137.4  V2A              0.2609            likely_benign


uniprot_id  protein_variant  am_pathogenicity  am_class

A0A024R1R8  M1A              0.4673            ambiguous

A0A024R1R8  M1C              0.3828            ambiguous

AlphaMissense_gene_hg19.tsv.gz, AlphaMissense_gene_hg38.tsv.gz

transcript_id      mean_am_pathogenicity

ENST00000000233.5  0.7422697635438503

ENST00000000412.3  0.37834258163288265

ENST00000001008.4  0.4222901115567318

ENST00000001146.2  0.4666058543393151


#CHROM  POS        REF    ALT       genome        transcript_id      protein_variant  am_pathogenicity  am_class

chr1    65568      A      C         hg38          ENST00000641515.2  K2Q              0.0938            likely_benign

chr1    65568      A      G         hg38          ENST00000641515.2  K2E              0.0766            likely_benign

chr1    65569      A      G         hg38          ENST00000641515.2  K2R              0.0756            likely_benign

chr1    65569      A      T         hg38          ENST00000641515.2  K2M              0.1732            likely_benign



transcript_id       protein_variant  am_pathogenicity  am_class

ENST00000000442.11  M1A              0.2808            likely_benign

ENST00000000442.11  M1C              0.1724            likely_benign

ENST00000000442.11  M1D              0.7278            likely_pathogenic

ENST00000000442.11  M1E              0.5328            ambiguous



Files (6.1 GB)

Name Size Download all
1.2 GB Download
243.9 kB Download
253.6 kB Download
622.3 MB Download
643.0 MB Download
2.5 GB Download
1.2 GB Download

Additional details