Google's new AI model could help explain why diseases develop
What's the story
Google DeepMind has unveiled a groundbreaking AI model, AlphaGenome 2, which seeks to understand how DNA changes affect different cells. The model is an upgrade from last year's genome-predicting AI, AlphaGenome. It promises to offer a more nuanced understanding of genetic variants' effects on individual cell types and states, potentially revealing disease-relevant effects in specific cells that might otherwise go unnoticed.
Enhanced prediction
Trained on over 100 million individual cell profiles
AlphaGenome 2 was trained on over 100 million individual cell profiles from 68 tissues, giving it a more detailed view than its predecessor.
Google DeepMind's genomics lead, Ziga Avsec, said this specificity is crucial for drug development as "you're targeting specific proteins and specific cell types."
The model identifies nearly twice as many variants that can change gene expression in cells than its previous state-of-the-art model.
Research tool
AlphaGenomes 2 can make predictions about new cell types
AlphaGenome 2 can also make predictions about new cell types and biological states without retraining.
It could identify gaps in experimental data, possibly guiding scientists on what to investigate next in the lab.
Adam Gayoso, a research scientist at Google DeepMind, said the big breakthrough is "tying human genetics to cell biology."
He added that AlphaGenome 2 gives researchers a way to explore genetic variants identified in large populations and connect those effects to disease.
Access and expansion
AlphaGenome 2 now available for non-commercial research
Google has made AlphaGenome 2 available for non-commercial research via the AlphaGenome API, with a limited "private preview" for commercial research.
DeepMind is now working on improving its genetics models by generating more data, including cell states not well represented today.