Google DeepMind releases AlphaGenome 2 to predict variant effects in single cells
Google DeepMind announced AlphaGenome 2, a DNA sequence model that predicts how genetic variants affect gene expression and other readouts inside specific human cell types and states, rather than averaged across bulk tissue. The model is conditional: alongside a million-letter DNA sequence it takes a target biological condition, either picked from a reference catalog or generated from a scientist's own single-cell experiment through a cell encoder. DeepMind says it was trained on more than 107 million individual cell profiles from human and mouse samples and identifies nearly twice as many expression-changing variants as the previous model. AlphaGenome 2 is available today for non-commercial research via the AlphaGenome API, with a private preview for commercial research.
Why it matters
Researchers can now predict variant effects in disease-relevant cell types that broad-tissue models average away, and can condition on their own single-cell data without retraining — which the source says can help them decide which experiments to run next.
Sources
- PrimaryGoogle DeepMind
Written by AIToolMaps from the source above and checked against it sentence by sentence.