Google’s AlphaGenome 2 zooms in on the cellular roots of disease
Summary
- Google DeepMind has unveiled AlphaGenome 2, a new AI model designed to predict how genetic variants affect individual cell types and states.
- The model was trained on over 100 million individual cell profiles across 68 tissues, identifying nearly twice as many gene-expression variants as its predecessor.
- AlphaGenome 2 can make predictions for new cell types and biological states without retraining and is available for non-commercial research via API.
- While offering high granularity for drug discovery and disease research, the model comes with a higher computational cost than the original AlphaGenome.
Why it matters
AlphaGenome 2 bridges human genetics and cell biology, giving researchers a predictive tool to understand disease mechanisms and target specific cell types for treatments.