Introducing AlphaGenome 2
On November 11, 2026, Google's influential AI lab, DeepMind, unveiled AlphaGenome 2, a powerful new model designed to predict how genetic variants affect the behavior of individual cells. While every cell in the body contains the same DNA, a skin cell functions
differently from a heart or immune cell because they switch different genes 'on' or 'off'. Most genetic variations linked to diseases like diabetes or heart disease are found in these regulatory 'switch' regions of DNA, not in the genes themselves. Understanding their impact has been incredibly difficult because a variant might be harmless in one cell type but disastrous in another. AlphaGenome 2 was built specifically to tackle this problem, providing a granular view that was previously impossible.
A Leap in Precision
The original AlphaGenome model analyzed genetic data at the broader 'tissue' level, which could dilute or miss signals from specific cells. AlphaGenome 2 changes the game by taking not just a DNA sequence but also a specific cell type or state as an input. For example, it can differentiate between a resting immune cell and one that has been activated by a vaccine. The results are dramatic. In tests, AlphaGenome 2 identified nearly twice as many genetic variants that alter gene expression compared to previous state-of-the-art models. On one immune cell dataset, it successfully identified the causal variant in a region 30% of the time, a significant jump from the 16% achieved by its predecessor.
From Theory to Real-World Disease
This isn't just a theoretical improvement; the model is already providing insights into real diseases. Researchers used it to study a genetic variant linked to severe cases of COVID-19. The AI revealed that the variant's disruptive effect on gene activity is much stronger in immune cells that have been 'activated' by a vaccine, a discovery that would be missed by looking at unactivated cells alone. In another example, AlphaGenome 2 was able to replicate a major discovery related to Inflammatory Bowel Disease (IBD) through computation alone. It successfully identified that a specific genetic change causes a gene called ETS2 to become overactive in certain immune cells within the gut, a conclusion that previously required extensive and complex laboratory experiments to prove.
The Bigger Picture: AI-Powered Drug Discovery
AlphaGenome 2 is the latest tool in Google's growing arsenal of biomedical AI, which aims to reshape drug discovery. It complements the work of its sister model, AlphaFold, which famously solved the problem of protein structure prediction and earned its creators a Nobel Prize. That technology now powers Isomorphic Labs, a Google spinoff dedicated to designing new medicines with AI. Isomorphic is already using AI to tackle targets once considered 'undruggable' and has multi-billion dollar partnership deals with pharmaceutical giants like Eli Lilly and Novartis. While AlphaFold helps design the 'key' (the drug), AlphaGenome 2 helps find the right 'lock' (the cellular mechanism driving the disease). By pinpointing which genetic variants are truly causal and in which cells they are active, the new model can help scientists focus their efforts, saving precious time and resources in the long journey to develop new treatments.
