A groundbreaking new artificial intelligence (AI) tool is revolutionizing the way we predict and prevent diseases. Delphi-2M, a large language model, has the ability to forecast a person’s risk of developing over 1,000 diseases based on their medical records and lifestyle factors. Developed by researchers and described in a study published in Nature, this tool can provide risk estimates up to 20 years in advance.
Unlike existing AI tools that focus on predicting a single disease, Delphi-2M is capable of modeling multiple diseases simultaneously. This capability is a game-changer in the medical field, as it allows clinicians to identify high-risk individuals and implement preventive measures early on. According to Stefan Feuerriegel, a computer scientist at the Ludwig Maximilian University of Munich, Delphi-2M’s ability to generate comprehensive health trajectories is “astonishing.”
The model is based on a modified large language model (LLM) called a generative pre-trained transformer (GPT), which is commonly used in AI chatbots. By training Delphi-2M on data from 400,000 participants of the UK Biobank, the researchers were able to accurately predict the likelihood of developing 1,258 diseases based on an individual’s medical history, age, sex, body mass index, and health-related habits.
Delphi-2M outperformed existing models in predicting the risk of developing multiple diseases and demonstrated high accuracy in forecasting conditions with predictable progression patterns, such as certain types of cancer. The model’s predictions were tested on health data from 1.9 million individuals in the Danish National Patient Registry, showing promising results even when applied to different national health systems.
While Delphi-2M represents a significant advancement in disease prediction, it does have limitations. For instance, the model relies on data from the UK Biobank, which may not capture the full medical history of participants. Researchers are working to evaluate the model’s accuracy on data sets from various countries to enhance its precision and scope.
In conclusion, Delphi-2M is a powerful tool that holds great potential for transforming healthcare by providing early warnings of disease and enabling proactive interventions. As researchers continue to refine and expand the model’s capabilities, the future of personalized medicine looks brighter than ever.
This article was originally published on September 17, 2025, in Nature magazine. If you’re interested in supporting impactful science journalism, consider subscribing to Scientific American to access more stories like this one.
