The short version

Google DeepMind announced AlphaGenome 2 on November 11, a research AI that predicts genetic sequence effects in specific human cell types.

A new approach to interpreting genetic variants

Google DeepMind says its AlphaGenome 2 model can predict how variations in a DNA sequence affect gene activity and other cell properties in hundreds of human cell types. Unlike an earlier version that issued broad predictions across many tissues at once, the new model conditions its analysis on a specified cell type or biological state. Its developers say the system was trained using more than 100 million single-cell profiles across 68 tissues and can help researchers focus on possible causes of disease.

Reported performance improvements still require external evaluation

DeepMind says its benchmark studies identify nearly twice as many expression-changing genetic variants as the earlier system. That is the company's reported research result, not an independently verified clinical benefit. Researchers will need to evaluate reproducibility, biases in training data, biological generalization and how model predictions compare with laboratory experiments. Genomic signals are complex, and a prediction should not be treated as proof of a disease cause.

Access and medical limitations

The team says AlphaGenome 2 is available for non-commercial research through an API and in private preview for commercial researchers. It is not validated or approved for clinical diagnostic use, nor does the announcement mean patient-specific treatment recommendations are available. CDN News will report peer-reviewed follow-up studies and real-world research uses without describing an exploratory model as a proven medical device.

Follow the evidence

Sources & transparency

Google DeepMind — AlphaGenome 2 announcement, November 11

Original CDN News coverage. Sources are linked for reference.

Source linked ·

Request a correction