MissenseDB

Name: MissenseDB
Published: 2026
License: https://creativecommons.org/licenses/by/4.0/

v2.0 · 2026

Junjie Xu, Stephanie Portelli, Aaron S. Kovacs, Ashar J. Malik and David B. Ascher

Proteome-wide missense variant effect predictions

Consolidated pathogenicity predictions from AlphaMissense, VESM, and VESPA across 19,436 human UniProtKB/Swiss-Prot proteins. Visualise susceptibility hotspots on 3D structures, explore per-residue substitution scores, cross-reference clinical significance from HUMSAVAR and dbSNP population frequencies, and query the database programmatically via an open REST API.

19,436 human proteins

10.7M residue positions

3 predictors

211,731 structure mappings

84,476 HUMSAVAR variants

Search by UniProt accession, gene symbol, or protein name e.g. P04637, TP53, cellular tumor antigen Autocomplete activates after 2 characters and matches across all fields. Selecting a result submits the UniProt accession — the unique stable identifier used internally to retrieve all precomputed scores, structural mappings, and variant annotations.

Three predictors

AlphaMissense, VESM, and VESPA scores are normalised to 0–1 and displayed side-by-side. Higher scores indicate greater predicted pathogenicity.

Structural context

Susceptibility scores are mapped to B-factors and rendered on experimental PDB structures and AlphaFold models using Mol* — hotspots light up directly on the 3D structure.

Clinical annotation

Variants are cross-referenced against HUMSAVAR (UniProt 2026_01) providing ACMG/AMP classifications, dbSNP population frequencies, ClinVar submissions, and disease associations where available.

Genomic coordinate lookup

Map GRCh38 nucleotide positions directly to protein residues and retrieve predictor scores in batch. Covers 18,292 proteins (94.1% of Swiss-Prot). Accepts VCF-style coordinates; returns UniProt accession, position, and optional predictor loads. Go to Genomic Lookup →