Google DeepMind Announces AlphaGenome 2 Model to Examine Genetic Variants

Google DeepMind researchers have introduced AlphaGenome 2, a new artificial intelligence model developed to examine how changes in DNA affect different cells.

As the successor to the original AlphaGenome, the new model was trained on over 100 million individual cell profiles spanning 68 tissues. Capable of detecting variants that alter gene expression nearly twice as effectively as its initial version, the model offers a more detailed perspective on genetic variants.

The model can make predictions about new cell types and biological states without requiring retraining, and it helps identify gaps in experimental data. According to DeepMind genomics lead Žiga Avsec, the model is of critical importance for drug development processes by targeting specific proteins and cell types.

While more costly than the original AlphaGenome model, AlphaGenome 2 became available for non-commercial research via the AlphaGenome API as of November 11, 2026, alongside the launch of a limited private preview period for commercial use.