Sarah Urbut: Predicting Your Health Arc

Key Points:

• Sarah Urbut and colleagues at Harvard Medical School developed ALADYNOULLI, an AI model integrating electronic medical records and polygenic risk scores from 683,000 individuals across three cohorts, to predict individual health trajectories over up to 52 years.
• ALADYNOULLI identifies 21 latent disease signatures and uses Gaussian processes to dynamically update risk predictions, outperforming standard clinical calculators like the pooled cohort equation for coronary artery disease and the GAIL model for breast cancer.
• The model reveals distinct genomic pathways underlying the same disease phenotypes, enabling insights into medication response and rare disease prediction, with applications across multiple conditions including heart attack, breast cancer, and depression.
• ALADYNOULLI complements other long-term health prediction models by incorporating diverse data types and longitudinal patient information, advancing personalized medical forecasting and enabling temporal risk estimation and counterfactual scenario analysis.
• The research highlights the growing potential of AI-driven medical forecasting to transform patient care by predicting not only disease risk but also timing, thereby supporting prevention and personalized intervention strategies.