Google DeepMind unveils AlphaGenome 2, which can predict the impact of genetic variants based on cell type.

Beating AI Express: Google DeepMind has announced the launch of its next-generation DNA sequence model, AlphaGenome 2, which predicts how DNA sequences and their variants affect specific human cell types to help study disease causes at the cellular level. Building on last year’s AlphaGenome, the model refines predictions from coarse tissue averages to individual cell types and states, with a redesigned architecture. Officials state that AlphaGenome 2 identifies nearly twice as many variants that alter cellular gene expression as the previous leading models. Trained on human and mouse population and single-cell data, its training set includes over 107 million single-cell atlases covering 68 tissues, turning static cell maps into predictive engines for human genetics. A conditional model, AlphaGenome 2 generates predictions for one tissue, cell type, or state at a time, adapts to new single-cell data from researchers without retraining, and imputes missing experimental data. The model is now open for non-commercial research via the AlphaGenome API; commercial research access is offered as a restricted non-public preview. Officials stress that AlphaGenome and AlphaGenome 2 are for research only, and have not been approved for clinical or diagnostic use.