Key Findings
Google DeepMind’s AlphaGenome 2 can predict how “non-coding” DNA sequences affect the expression of other genes across 637 types of human cells and 311 kinds of mouse cells. In a validation case, the model replicated findings related to a DNA variant in the ETS2 gene—a driver of inflammatory bowel disease (IBD)—in approximately 30 minutes. This specific discovery had previously taken researchers at the Francis Crick Institute a year and a half of experimental work using gene-editing and molecular-biology techniques.
Technical Details
AlphaGenome 2 improves upon its predecessor, AlphaGenome 1, by addressing the lack of cell-type specificity. While AlphaGenome 1 was trained on “bulk” samples of heterogeneous cells, AlphaGenome 2 was trained on measurements from more than 100 million cells across 68 tissues, utilizing single-cell gene activity data. This approach allows the model to provide a more fine-grained representation of cellular differences. Additionally, the model is flexible; researchers can input new single-cell gene-activity data, such as data from cells in a disease state, to generate specific gene-variant predictions.
Background & Context
Non-coding DNA sequences comprise 98% of the human genome and are among the most difficult to decipher. Many of these sequences, including gene enhancers, are active only in specific cell types. Previously, Google DeepMind released the AlphaGenome Atlas, which provided predictions for nine billion possible single-letter DNA changes to the human genome, but these predictions did not specify which cell types—such as pancreatic β-cells or B cells—were affected.
Significance & Outlook
The ability to predict variant effects in diverse cells is critical for understanding genetic diseases. By eliminating the need for immediate physical experimentation (described as “lifting a pipette”), AlphaGenome 2 allows researchers to identify disease-underlying variants more rapidly. This capability streamlines the process of mapping how variants in “gene deserts” influence protein-coding sequences and disease pathology.
Source: https://www.nature.com/articles/d41586-026-03673-0
Get our weekly technology intelligence — free
Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.
Subscribe Free — Weekly Tech Intelligence
By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.
- Your email and selected fields are used only to deliver the newsletter.
- We never share your information with third parties.
- You can unsubscribe anytime via the link in each email.
See our Privacy Policy for details.
Takes about a minute · Unsubscribe anytime



