Google’s new AI reveals hidden genetic clues in new immune-cell test
Google DeepMind announced AlphaGenome 2, a DNA prediction model that conditions its predictions on a target tissue, cell type or state. Trained on bulk and single-cell data from human and mouse samples, it aims to identify genetic effects that broad tissue-level measurements may miss and help researchers prioritize experiments. DeepMind reports that it identified 30% of causal variants on an immune-cell benchmark, compared with 16% for the original model; these results do not measure diagnostic accuracy or personal disease risk. The company also says the model computationally reproduced an established bowel-disease mechanism involving a regulatory variant that increases ETS2 activity in gut macrophages. AlphaGenome 2 is available via API for non-commercial research, with commercial research in private preview, and is not validated for clinical or diagnostic use.
