AlphaGenome 2: Google’s New AI Gets Better at Decoding Disease-Linked DNA

Google DeepMind’s AlphaGenome 2 improves predictions of disease-linked genetic variants using 107 million cell profiles, with potential benefits for drug research.

Verfasst von
Aminu Abdullahi
Aminu Abdullahi
Nov 12, 2026
AlphaGenome 2: Google’s New AI Gets Better at Decoding Disease-Linked DNA

Google DeepMind launches AlphaGenome 2. Image: Sangharsh Lohakare/Unsplash

Every cell in your body carries the same DNA blueprint, yet a gut macrophage reacts to genetic glitches completely differently than a heart muscle cell.

Google DeepMind announced AlphaGenome 2 on Nov. 11, introducing a DNA analysis model that predicts how genetic variations influence individual cell types and states.

Trained on more than 107 million single-cell profiles across 68 tissues, the model nearly doubled its predecessor’s performance on a benchmark for identifying disease-linked causal variants, according to the company.

The improvement could help researchers narrow down genetic drivers of disease and identify promising drug targets before moving into costly laboratory experiments.

The original AlphaGenome, previewed in June 2025 and published in Nature in January 2026, predicted molecular activity across thousands of tissue experiments simultaneously, an approach DeepMind said “spread the model’s capacity thin.” AlphaGenome 2 replaces that parallel design with a conditional architecture.

Advertisement

It takes up to one million DNA letters alongside one specific biological condition—a tissue, cell type, or cell state—and evaluates them together at every stage.

This allows researchers to capture subtle regulatory shifts, including newly added DNA methylation readouts alongside gene expression, histone modifications, and chromatin accessibility.

Metric/FeatureOriginal AlphaGenomeAlphaGenome 2
Causal Variants Identified (TenK10K)16%30%
Directional Gene Effects Predicted23%43%
Prediction ArchitectureParallel bulk tissue outputsSingle-condition conditional model
Training ScopeBulk tissue and cell lines107M+ single-cell profiles across 68 tissues

“What makes AlphaGenome 2 exciting is how it adapts directly to new single-cell data without retraining, predicting how DNA variants act in newly profiled cell states from even a small sample of cells,” said Sarah Teichmann, co-founder of the Human Cell Atlas consortium and professor at the University of Cambridge, in Google DeepMind’s announcement.

Real-world validation and compute reality

In testing, DeepMind demonstrated the model’s ability to pinpoint disease mechanisms without lab-heavy trial and error:

  • Inflammatory Bowel Disease (IBD): AlphaGenome 2 computationally reproduced a landmark study showing how a distant regulatory variant overactivates the ETS2 gene specifically inside gut macrophages.
  • COVID-19 Immune Response: Using unseen mRNA vaccine data, the model separated activated T cells from resting ones, showing that a severe COVID-19 variant disrupts gene activity far more intensely in activated cells.
Advertisement

However, higher cellular resolution demands heavier compute. Genomics lead Žiga Avsec told The Verge that AlphaGenome 2’s detailed predictions are “more expensive” to generate, meaning it will coexist with—rather than replace—the original model used for massive, genome-wide scans.

How AlphaGenome 2 affects researchers and drug developers

For biotech firms, pharmaceutical teams, and academic labs, AlphaGenome 2 acts as a high-speed triage engine before costly wet-lab experiments begin. Because drug discovery relies on targeting exact proteins inside specific cells, pinpointing whether a variant turns a gene up or down narrows down therapeutic candidates faster.

“When you develop drugs, you’re targeting specific proteins and specific cell types,” Avsec told The Verge.

Academic researchers can access AlphaGenome 2 for free through the AlphaGenome API, while commercial organizations can apply for a private preview. DeepMind also plans to introduce AlphaGenome 2 Skills for its Google Antigravity platform.

The model remains intended for non-clinical research and is not approved for diagnostic use. Its predictions still require experimental validation before they can inform therapeutic decisions.

Advertisement

For pharmaceutical companies and research laboratories, the immediate opportunity is to narrow down which genetic variants deserve closer investigation. Whether those predictions translate into faster drug development will depend on how well they hold up in laboratory testing.

Other news: Google is expanding its Gemini Intelligence features beyond its own Pixel lineup, bringing more personalized AI assistance to the Pixel 10 series and Motorola Signature 27 as it pushes to make smartphones more context-aware.

Aminu Abdullahi

Aminu Abdullahi is a B2C and B2B technology and finance writer with more than six years of experience covering enterprise IT, cybersecurity, cloud computing, artificial intelligence, fintech, business software, and emerging technologies. He has written for a wide range of technical and business audiences, from IT professionals and cybersecurity leaders to small business owners, executives, and technology buyers. His work has appeared in publications including: TechRepublic eWEEK Channel Insider Geekflare Enterprise Networking Planet eSecurity Planet CIO Insight Webopedia With a background in computer science, Aminu specializes in translating complex technical subjects into clear, practical, and accessible content. His writing helps readers understand emerging technologies, evaluate business software, strengthen cybersecurity strategies, and make more informed decisions about technology investments. Across his work, Aminu focuses on the real-world impact of technology, connecting technical innovation with business value, operational efficiency, security, and long-term digital transformation.