Overview
- Google DeepMind released AlphaGenome 2 on Wednesday, Nov. 11, 2026, and made the model available for non-commercial research via the AlphaGenome API with a private commercial preview.
- The model was trained on more than 100 million single-cell profiles across 68 tissues from human and mouse samples to predict how DNA variants alter molecular activity in specific cell types and states.
- DeepMind reports internal benchmarks showing AlphaGenome 2 finds nearly twice as many expression-altering variants in cells compared with prior state-of-the-art models.
- AlphaGenome 2 accepts long DNA stretches plus a target cellular condition and can impute unmeasured molecular data such as chromatin accessibility or DNA methylation to help researchers decide which wet-lab experiments to run next.
- The tool could speed how scientists pinpoint disease-linked variants and plan experiments, but its higher computational cost and the need for independent validation will shape how quickly labs and companies adopt it.