For decades, scientists have struggled to understand what most of our DNA actually does. While the Human Genome Project mapped our complete genetic code, the function of 98% of it—the non-coding regions that don't directly produce proteins—remained largely mysterious.
On June 25, 2025, Google DeepMind unveiled AlphaGenome, an artificial intelligence system designed to illuminate this genomic 'dark matter.' The model can process DNA sequences up to one million letters long and predict thousands of molecular properties, including gene expression levels, RNA splicing patterns, and the impact of mutations across different cell types and tissues.
"This is one of the most fundamental problems not just in biology—in all of science," said Pushmeet Kohli, DeepMind's head of AI for science. The model represents a unified approach to genome interpretation, combining convolutional neural networks to detect short patterns with transformers to model long-range interactions.
In rigorous testing, AlphaGenome outperformed specialized tools in 24 out of 26 variant effect prediction tasks. When applied to leukemia research, it accurately predicted how non-coding mutations activate cancer-driving genes—a capability previously requiring extensive laboratory experiments.
"For the first time, we have a single model that unifies long-range context, base-level precision, and state-of-the-art performance across a whole spectrum of genomic tasks," noted Dr. Caleb Lareau of Memorial Sloan Kettering Cancer Center, who had early access to the tool.
While still in its early stages, AlphaGenome could accelerate disease research by helping scientists identify which genetic variants cause conditions, potentially revolutionizing personalized medicine. DeepMind has made the model available via API for non-commercial research and plans a full release in the future. According to Demis Hassabis, DeepMind's CEO, this represents a step toward his dream of creating a "virtual cell" for drug studies and medical research.