AlphaGenome

Google DeepMind has released AlphaGenome 2, shifting DNA variant effect prediction from tissue-level averages to individual cell types and states, identifying nearly twice as many expression-altering variants as the previous best model. Built on a conditional architecture trained on over 107 million single-cell profiles with cross-modal imputation support, it is now available via API for non-commercial research.

Google DeepMind has released AlphaGenome 2, a successor to its genome-predicting model that forecasts how genetic variants affect individual cell types and states. The model targets the challenge that the same variant can have divergent effects across cells, offering finer resolution for disease mechanism research and therapeutic discovery.