Google’s New AI Reads DNA to Uncover Hidden Disease Risks

AlphaGenome operates as a virtual laboratory, allowing scientists to simulate the effects of mutations without expensive and time-consuming wet lab experiments.

Google DeepMind has unveiled AlphaGenome, an AI capable of scanning long stretches of DNA to predict how minor genetic variations influence gene activity, offering new insights into diseases from cancer to heart and mental health conditions.

Scientists have long struggled to interpret much of the human genome, as only about two per cent of DNA codes for proteins. The remaining 98 per cent, once dismissed as “junk DNA,” functions as a regulatory system controlling when, where, and how genes are activated.

Many disease-associated mutations lie in this regulatory “dark genome,” making it difficult to identify which variants drive disease. Google’s AlphaGenome has now bridged this gap, enabling a deeper understanding of how genetic changes translate into biological effects.

Detailed in the Nature study Advancing regulatory variant effect prediction with AlphaGenome, the AI tool reads up to one million DNA letters at a time with single-letter precision, predicting how genetic variants influence gene expression, splicing, chromatin structure, and three-dimensional DNA folding.

This ability allows researchers to see how a single mutation can ripple across multiple molecular processes, potentially triggering conditions such as cancer, autoimmune disorders, psychiatric illnesses, or heart disease.

Unlike earlier models that could analyze either long DNA sequences with low detail or short sequences at high resolution, AlphaGenome combines both, offering a comprehensive view of gene regulation.

In testing, the tool has matched or outperformed existing models in 25 of 26 benchmark tasks, accurately reproducing known disease mechanisms. For example, it successfully identified regulatory changes near the TAL1 oncogene linked to T-cell leukaemia and predicted splicing errors that underlie rare genetic disorders.

AlphaGenome operates as a virtual laboratory, allowing scientists to simulate the effects of mutations without expensive and time-consuming wet lab experiments. It can help prioritize which genetic variants are likely disease-causing, design gene therapies targeted to specific tissues, and improve risk prediction by clarifying which mutations truly matter.

“We believe AlphaGenome can be a valuable resource for the scientific community, helping scientists better understand genome function, disease biology, and ultimately, drive new biological discoveries and the development of new treatments,” said Google DeepMind.

Experts emphasize that while the tool is transformative, its predictions still require experimental validation, as human cells may behave differently than models anticipate.

Ben Lehner of the Wellcome Sanger Institute notes that data quality is a key limitation: “Most existing data in biology is not very suitable for AI, the datasets are too small and not well standardized.”

Currently available for non-commercial research, AlphaGenome is already being used by thousands of scientists worldwide to decode the regulatory genome and explore the genetic roots of complex disease, signalling a step closer to precision medicine powered by artificial intelligence.

Stay tuned for more such updates on Digital Health News