DeepMind Unveils AlphaGenome 2: Predict DNA Variant Effects With Cell Type-Level Precision

Google DeepMind has released AlphaGenome 2. This is a conditional DNA sequence model built to predict regulatory activity and variant effects, and delivers results segmented by individual cell type. This piece is sourced from MarkTechPost, authored by Asif Razzaq, and published on November 11, 2026.

A large portion of variants in DNA are not located in protein-coding regions. They fall in regulatory regions, influencing when and in which cell types genes get expressed. The exact same variant can produce completely different effects across different cell types. This is exactly the problem AlphaGenome 2 is designed to address: predicting variant effects in the explicit context of specific cell types.

As stated in the original post, the model can predict regulatory activity and variant effects per cell type. To elaborate, given an input DNA sequence, the model outputs corresponding regulatory activity, as well as the impact of variants on that regulatory activity, while distinguishing results across different cell types. This brings direct practical value for research on non-coding variants, interpretation of GWAS signals, and understanding of cell type-specific regulatory mechanisms.

In terms of availability, AlphaGenome 2 is provided through the AlphaGenome API for non-commercial research use. The official announcement did not release details on model performance, training data scale, the number of covered cell types, nor the improvement magnitude compared to the previous generation. This information will be released in follow-up papers or technical documentation.

Original article link: https://www.marktechpost.com/2026/11/11/google-deepmind-alphagenome-2-cell-type-variant-effects/