A new AI model from Google DeepMind, called AlphaGenome, promises to change how scientists understand DNA, the blueprint of the human body. It could reveal why small differences in our genetic code make some people more prone to conditions like high blood pressure, obesity, and dementia. The technology may also speed up research into genetic disorders and cancer.
Exploring the Dark Genome
While only 2% of DNA codes for proteins, the remaining 98%—known as the dark genome—controls how genes work. Mutations in this region often trigger diseases, but their exact role has been hard to study. AlphaGenome can analyze large sections of DNA at once, predicting both gene locations and how the dark genome influences gene expression and protein production.
Predicting Genetic Mutations
The AI can even predict how changing a single DNA letter may affect health. Natasha Latysheva, a DeepMind research engineer, explains that AlphaGenome helps identify mutations that cause rare genetic disorders and could guide new drug development. Researchers hope the AI will also assist in synthetic biology and designing DNA sequences for gene therapies.
Advancing Obesity, Diabetes, and Cancer Research
Scientists like Dr. Gareth Hawkes from the University of Exeter are using AlphaGenome to study obesity and diabetes. Many gene variants linked to these conditions are in the dark genome, and AlphaGenome predicts their effects, helping researchers test hypotheses in the lab. Similarly, the AI can identify mutations fueling cancer and determine which ones are potential drug targets.
A Milestone in Genomic AI
Experts describe AlphaGenome as a major breakthrough. Dr. Robert Goldstone from the Francis Crick Institute calls it “an incredible technical feat” for predicting gene activity from DNA alone. Prof. Ben Lehner of the Wellcome Sanger Institute notes the model performs well across hundreds of thousands of experiments, though it still needs refinement, especially for long-distance gene regulation and tissue-specific accuracy.
How AlphaGenome Works
Unlike language models like ChatGPT, AlphaGenome is a sequence-to-function model, analyzing how changes in DNA affect biological outcomes. It was trained using human and mouse cell experiment data, and over 3,000 scientists have already used it for research.
The Future of AI in Genetics
DeepMind’s success with AlphaFold, which predicts protein structures, won the team a Nobel Prize in Chemistry in 2024. AlphaGenome builds on this, potentially driving breakthroughs in medicine, drug discovery, and synthetic biology. As Pushmeet Kohli from DeepMind says, AI could usher in a new era of scientific discovery.
