ALADYNOULLI Beats Clinical Risk Tools Across 348 Diseases in 683,000-Patient Study
Updated
Updated · erictopol.substack.com · Aug 4
ALADYNOULLI Beats Clinical Risk Tools Across 348 Diseases in 683,000-Patient Study
3 articles · Updated · erictopol.substack.com · Aug 4
Summary
Nature published results showing ALADYNOULLI predicted individual 1- and 10-year disease risk across 348 diseases using data from 683,000 people in three cohorts followed for up to 52 years.
The model combines electronic medical records with 36 polygenic risk scores and maps each patient into 21 latent disease signatures, letting risk update over time rather than relying on static clinical snapshots.
Against standard tools, ALADYNOULLI posted markedly better 1-year discrimination for coronary artery disease—0.89 versus 0.68 for the pooled cohort equation—and for breast cancer—0.783 versus 0.54 for the GAIL model.
Those signatures also linked the same disease to different biological pathways, helping flag likely medication failure such as SSRI nonresponse in depression and improving prediction of rare diseases.
The work adds to a fast-growing wave of medical forecasting models, suggesting longer, richer longitudinal and multimodal patient data could sharpen prevention and timing of disease onset.