❯ Google launches AlphaGenome 2 to predict the effects of DNA variants in specific cells

A closer view of cellsGoogle DeepMind released AlphaGenome 2 on November 11 to predict how DNA variants affect particular cell types and states. Trained on more than 100 million individual cell profiles across 68 tissues, it helps researchers ask why the same genetic change affects one kind of immune cell much more than another.

Beyond tissue averagesThe first AlphaGenome predicted molecular activity across broader tissues, which can contain many different cell types. The new model takes a target biological condition alongside the DNA sequence and can use small samples of newly measured cells to predict unfamiliar states without retraining. This may reveal effects that are hidden when researchers average across an entire tissue.

Improvements in research testsDeepMind reports that, on the TenK10K dataset covering 28 primary immune-cell types, AlphaGenome 2 identified 30% of causal gene-activity variants versus 16% for its predecessor adapted to single-cell data. It also computationally reproduced a known inflammatory bowel disease mechanism involving increased ETS2 activity in gut macrophages. Such predictions can narrow experimental priorities, while biological mechanisms still require laboratory validation.

Research access opensThe API and tutorials are available for noncommercial research, with commercial research in private preview. The model has not been validated or approved for clinical or diagnostic use. Its immediate role is to help researchers prioritize variants, genes and cells for investigation, rather than supply ready-to-use treatments.

▮ SIGNALIdentifying the cells in which a genetic variant matters could reduce unproductive early experiments and make subsequent drug research more targeted.