Google DeepMind's AlphaGenome 2 reads DNA cell by cell and finds disease-causing variants twice as often
Trained on more than 107 million single-cell profiles, it predicts how a DNA change affects genes in a specific tissue.
- Training data: 107M+ cell profiles
- Variants found: 30% (was 16%)
- Direction correct: 43% (was 23%)
What it does
AlphaGenome 2, announced on 11 November, predicts how changes in DNA affect gene activity in one particular cell type or tissue at a time. On an immune-cell test, it found the variants behind effects in 30% of cases, up from 16% for the first version, and got the direction of the effect right in 43% of cases versus 23%. It can also predict DNA methylation, a chemical tag that switches genes on or off.
Who can use it
It reproduced a known finding linking a variant in the ETS2 gene to inflammatory bowel disease. It is free for non-commercial research through the AlphaGenome API, with commercial use in private preview. DeepMind says it is not validated for clinical or diagnostic use.






