Newsletters
European researchers have introduced Delphi-2M, an innovative artificial intelligence system capable of estimating a person’s risk for over 1,000 diseases, including cancer, diabetes, heart disease, and sepsis, potentially decades before symptoms appear.
Developed at the European Molecular Biology Laboratory (EMBL) in Cambridge, Delphi-2M uses a generative AI architecture similar to large language models but customized for health care. The system was trained on anonymized records from approximately 400,000 U.K. Biobank participants and validated on 1.9 million patient records from Denmark’s National Patient Registry.
Unlike conventional risk calculators that focus on specific conditions, Delphi-2M provides a broad view of a person’s long-term health trajectory. In tests, the AI achieved prediction accuracy comparable to established tools, such as the QRisk score for heart conditions.
A standout feature of Delphi-2M is its robustness across different datasets. Even though it was trained on U.K. records, it performed effectively when applied to Danish patient data, highlighting its potential for broader public health use.
While highly promising, Delphi-2M is not yet ready for routine clinical application. Researchers plan to integrate genetic and protein data in the future to further refine its predictive abilities.
The team, including scientists from the German Cancer Research Center, has patented key innovations and is exploring commercialization opportunities. Experts believe Delphi-2M represents a significant leap forward in preventive medicine, paving the way for highly personalized health care.
9 min read
