Genetic scores add more predictive power at lower clinical risk

In an observational preprint analyzing 900,000 human participants from UK Biobank and FinnGen, researchers developed and validated models to predict 150 diseases and all-cause mortality. Genetic information provided the largest predictive gains in metabolic, cardiovascular, autoimmune, and neurological conditions, but contributed little to predicting genitourinary and respiratory diseases. Across diseases, all significant interactions between polygenic risk scores and clinical risk were negative, showing that genetic data delivered larger predictive power gains when clinical risk was lower. The authors suggest polygenic scores can reveal disease susceptibility not yet clinically apparent. These clinical-genetic models generalized across three biobanks.