AI tool forecasts risk of more than 1,000 diseases up to two decades ahead

A new artificial intelligence tool can forecast the likelihood and timing of more than 1,000 diseases, including cancer, up to 20 years into the future, according to researchers. The system, called Delphi-2M, was described in Nature by teams from the European Molecular Biology Laboratory, the German Cancer Research Centre and the University of Copenhagen.

It analyzes anonymized medical histories, 'medical events' and lifestyle factors such as smoking, alcohol consumption and weight. It was trained and tested on data from about 400,000 UK Biobank participants and 1.9 million Danes in the national patient registry. Tomas Fitzgerald, an expert in molecular biology and study co-author, said: "Medical events often follow predictable patterns. Our AI model learns those patterns and can forecast future health outcomes."

In clinical practice, experts envision use in general practitioners' offices to identify high-risk patients and tailor preventive advice. Professor Ewan Birney, executive director of the EMBL, described a scenario in which a clinician could say, "Here's four major risks that are in your future and here's two things you could do to really change that." He added that weight loss and smoking cessation would likely be emphasized, and that such guidance would be reflected in a patient’s data.

Moritz Gerstung, a computational cancer biologist at the German Cancer Research Centre, said: "This is the beginning of a new way to understand human health and disease progression." He noted the model could also help inform disease-screening programmes and anticipate demand in healthcare, such as estimating how many people in a city could suffer a heart attack in the coming year.

Still, some experts urged caution. Justin Stebbing, an oncologist at Anglia Ruskin University, warned against interpreting the results as direct causal relationships, noting that the model reproduces biases found in its training data, including healthy volunteer and selection biases. Peter Bannister, a fellow at the Institution of Engineering and Technology, added that the datasets are biased in terms of age, ethnicity and current health outcomes, and stressed the need for robust digital infrastructure and broad access so the technology benefits those most in need.

Gustavo Sudre, an expert in neuroimaging and AI at King’s College London, said the current version relies on anonymized clinical records but highlighted that the architecture was designed to accommodate richer data types, such as biomarkers, imaging and genomics. He argued that with future integrations, the Delphi platform could evolve into a multimodal, precision-medicine tool.

The research arrives as health advocates warn that cancer cases are projected to reach record highs by 2040, with forecasts suggesting a new case every two minutes in some regions. The most common cancers—breast, prostate and lung—are expected to remain prevalent even as incidence climbs. Researchers acknowledge that Delphi-2M aims to forecast multiple diseases over long time horizons and alongside routine care, but they caution that substantial work remains to ensure clinical validity, equitable deployment and that benefits reach all populations.