Static

Google’s AlphaGenome 2 zooms in on the cellular roots of disease

First reported by The Verge ·

The signal ●○○○ Compiled by AI from The Verge, the single source so far
Why you might care

Drug discovery targeting specific cell types becomes more precise with this advanced AI, potentially accelerating the development of treatments for diseases.

What happened

Google DeepMind has launched AlphaGenome 2, an advanced AI model designed to understand how DNA variations impact specific cell types and states. This successor to AlphaGenome was trained on over 100 million individual cell profiles across 68 tissues, enabling it to provide a more detailed view of genetic variants than its predecessor. The model can predict the effects of genetic variants on individual cells, potentially revealing disease-specific cellular impacts that might otherwise be missed. A key advancement is its ability to make predictions about new cell types and biological states without retraining, and it can also identify gaps in experimental data. AlphaGenome 2 identifies nearly twice as many variants that change gene expression in cells compared to previous state-of-the-art models. The researchers highlighted that this specificity is crucial for targeted drug development, which often focuses on particular proteins and cell types. AlphaGenome 2 is now available for non-commercial research via API, with a limited commercial preview.

What it means

AlphaGenome 2 represents a significant leap in the application of AI to genetics, moving beyond broad genome-wide analysis to provide cell-type-specific predictions. This granularity allows researchers to connect genetic predispositions to specific cellular mechanisms, a critical step in understanding and treating complex diseases. The model's ability to generalize to new cell types and identify experimental gaps suggests a future where AI can proactively guide scientific inquiry, making research more efficient and targeted.

This development signals a growing trend of AI-driven precision in biological research, with Google DeepMind positioning itself at the forefront of translating complex biological data into actionable insights. The model's capabilities could profoundly affect pharmaceutical companies and academic researchers by refining drug target identification and potentially reducing the time and cost associated with preclinical research. The ongoing efforts by various AI labs to tackle scientific challenges indicate an acceleration of AI's role in scientific discovery across multiple domains.

AI-written summary. May contain errors.