STRC Research

AlphaMissense

2 min read

AlphaMissense

TL;DR: DeepMind’s missense-pathogenicity predictor (2023). Every STRC amino-acid substitution has a score; E1659A scores 0.9016 — high-confidence pathogenic. Used as ACMG PP3 supporting evidence.

Google DeepMind’s missense variant pathogenicity predictor. Classifies all possible single amino acid changes across the human proteome as likely benign, ambiguous, or likely pathogenic.

What It Does

  • Predicts pathogenicity of missense variants (amino acid substitutions)
  • Score 0-1: <0.34 likely benign, 0.34-0.564 ambiguous, >0.564 likely pathogenic
  • Pre-computed for ALL possible missense variants in human proteome (71M variants)
  • Based on protein language model + structural features

How to Use

Web (via AlphaFold DB)

  1. Go to https://alphafold.ebi.ac.uk/entry/Q7RTU9 (STRC)
  2. Click “AlphaMissense” tab
  3. View heatmap of all possible substitutions
  4. Download full CSV

Download Pre-computed Data

# Full dataset (4.3GB)
gsutil cp gs://dm_alphamissense/AlphaMissense_hg38.tsv.gz .
 
# Or per-gene from AlphaFold DB API
curl "https://alphafold.ebi.ac.uk/api/prediction/Q7RTU9" | python3 -c "
import sys,json
d = json.load(sys.stdin)[0]
print(d.get('amAnnotationUrl', 'N/A'))
"

Python (from downloaded TSV)

import pandas as pd
am = pd.read_csv("AlphaMissense_hg38.tsv.gz", sep='\t', comment='#')
strc = am[am['uniprot_id'] == 'Q7RTU9']
e1659a = strc[(strc['protein_variant'] == 'E1659A')]
print(e1659a)  # Score: 0.9016

Verified Status

VERIFIED — STRC E1659A score: 0.9016 (likely pathogenic, ≥0.840 threshold per Pejaver 2022). Used as PP3_Moderate evidence in ACMG classification.

STRC Research Usage

Results (April 2026)

  • Full STRC heatmap analyzed: 33,725 variants scored. E1659 region (1651-1700) is the most pathogenic-sensitive area in STRC — 69% of substitutions score as pathogenic.
  • Regional pathogenicity: C-terminal (1476-1775): 43.1% pathogenic. mini-STRC region (700-1775): 39.6%. N-terminal (1-699): 27.1%.
  • Still untapped: REVEL cross-validation at discordant positions, STRCP1-specific variant analysis

Connections

Graph View

Backlinks