Photo: depositphotos
Google DeepMind has announced a new artificial intelligence tool, AlphaGenome, designed to study genetic diseases. The development aims to help scientists understand how mutations affect gene function. In the long term, this could open the way to new treatment methods, The Guardian reports.
Researchers at Google DeepMind said AlphaGenome will help identify genetic factors behind diseases and speed up the development of therapies. The tool predicts how mutations alter gene regulation, including where and in which cells genes are activated. Most common inherited diseases—such as cardiovascular conditions, autoimmune disorders, and mental health conditions—are linked specifically to regulatory mutations. The same mechanisms underlie many types of cancer, although identifying the key disruptions has so far been difficult.
“We see AlphaGenome as a tool for understanding what the functional elements of the genome do, and we hope this will accelerate our fundamental understanding of the code of life,” said DeepMind researcher Natasha Latysheva during a briefing.
The human genome consists of approximately 3 billion pairs of DNA letters. Only about 2% of this code is responsible for producing proteins, while the rest controls when, where, and how strongly genes are expressed.
AlphaGenome was trained on public genetic databases from humans and mice. The AI can simultaneously analyze up to one million DNA letters and predict the effects of mutations on various biological processes. DeepMind believes the tool will help more accurately identify critical regions of the genome involved in tissue development, including neural and liver tissues. AlphaGenome could also become a foundation for new gene therapies and for designing artificial DNA sequences with specific properties.
Karl De Boor of the University of British Columbia, who was not involved in the development, noted: “AlphaGenome can determine whether mutations affect genome regulation, which genes are impacted and how, and in which cell types.” According to him, this opens the possibility for targeted drug development.
“Ultimately, our goal is to have models that are so good we won’t need to run experiments to validate their predictions. While AlphaGenome is a significant innovation, achieving this goal will require sustained effort from the scientific community,” the developers said.
Some scientists have already begun using AlphaGenome in their research. Mark Mansour of the University of London called the tool “a significant step in changing” approaches to identifying the genetic causes of cancer.
“The non-coding genome makes up 98% of our genome’s 3 billion base pairs, and the ability to make predictions about this region is a major step forward,” emphasized statistical geneticist Gareth Howells of the University of Exeter.
At the same time, the handling of health data by other Google AI tools has sparked strong criticism within the scientific community. Studies have identified cases of medical misinformation being spread through AI Overviews. In one instance, Google provided incorrect information about liver function tests, potentially misleading users. Another study showed that Google AI Overviews, when responding to medical queries, most often cited YouTube while ignoring specialized medical resources. An analysis of more than 50,000 queries in Germany found that videos had become the primary information source, surpassing professional reference materials and hospital websites.
