European Molecular Biology Laboratory and German Cancer Research Center researchers have developed Delphi-2M, a generative transformer AI model trained on UK Biobank and Danish registry data from 400,000+ individuals, predicting 20-year risks for 1,258 diseases with accuracy rivaling single-disease tools. Integrating demographics, lifestyle (BMI, smoking, alcohol), and medical history, it excels for predictable conditions like cancers and myocardial infarction but varies for heterogeneous ones. Validated on 1.9 million external records, Delphi-2M generates synthetic trajectories for privacy-preserving research and highlights co-morbidity dynamics, potentially enabling early interventions and reducing screening costs by 60%. Rollouts in NHS pilots are planned for 2026, marking a shift to multi-disease forecasting in precision medicine.
Source: Nature & Medscape
Leave a Reply