Skin, Decoded
← Contents

Reference

Sources

Every claim traces to one of these

Advancing regulatory variant effect prediction with AlphaGenomeAvsec Ž. et al., Nature 649, 1206–1218 (2026)

The peer-reviewed AlphaGenome 1 paper.

AlphaGenome 2: Deciphering how cells read DNA to better understand diseaseGoogle DeepMind, 11 November 2026

Announcement, architecture, benchmark figures and limitations.

AlphaGenome 2 preprintbioRxiv, November 2026 · doi:10.64898/2026.11.11.765857

Not yet peer-reviewed.

AlphaGenome Atlas: in silico mutagenesis of the entire human genomeCheng J. et al., medRxiv 2026 · doi:10.64898/2026.09.16.26363192

Pre-computed scores for all ~9 billion single-letter changes.

AlphaGenome Terms of Service and documentationGoogle DeepMind

Licensing: non-commercial research use, no clinical decision-making, no training other models.

A disease-associated gene desert directs macrophage inflammation through ETS2Stankey C.T. et al., Nature 630, 447–456 (2024)

The chr21q22 / ETS2 / IBD finding reproduced by AlphaGenome 2.

Common loss-of-function variants of the epidermal barrier protein filaggrin are a major predisposing factor for atopic dermatitisPalmer C.N.A. et al., Nature Genetics 38, 441–446 (2006)

The founding filaggrin–eczema association.

Variants of the melanocyte-stimulating hormone receptor gene are associated with red hair and fair skin in humansValverde P. et al., Nature Genetics 11, 328–330 (1995)

MC1R and pigment phenotype.

Filaggrin mutations associated with skin and allergic diseasesIrvine A.D., McLean W.H.I., and Leung D.Y.M., New England Journal of Medicine 365, 1315–1327 (2011)

Review; carrier frequency and barrier mechanism.

Licensing note. AlphaGenome and AlphaGenome 2 are made available by Google DeepMind for non-commercial research use. Their outputs must not be used for clinical decision-making or to train other machine-learning models. This guide explains published findings; it does not call the model or reproduce its outputs.