Google DeepMind introduced AlphaGenome 2, trained on roughly 100 million cell profiles from 68 tissues. It predicts how DNA variants affect gene regulation across cell types and states using contexts of up to one million bases. In the company’s evaluations it explained nearly twice as many variants as its predecessor. A non-commercial research API and private commercial preview are available. Those figures describe specific benchmarks, not validated patient diagnosis or treatment outcomes.
Back to Tech
AlphaGenome 2 predicts variant effects across cell types and states
The research model uses a million-base DNA context; it is not a clinical diagnostic tool.