Abstract
Obesity and type 2 diabetes (T2D) are metabolic diseases with shared pathophysiology. Traditional polygenic risk scores (PRSs) have focused on these conditions individually, yet the single-disease approach falls short in capturing the full dimension of metabolic dysfunction. We derived a biologically enriched metabolic PRS (MetPRS), a composite score that uses multi-ancestry genome-wide association studies of 20 metabolic traits from over 8.5 million individuals. MetPRS, optimized to predict obesity (O-MetPRS) and T2D (D-MetPRS), outperformed existing PRSs in predicting obesity and T2D across six ancestries. O-MetPRS and D-MetPRS effectively identify individuals at high risk for metabolic multimorbidity and predict clinical outcomes, including GLP-1 receptor agonist initiation. O-MetPRS and D-MetPRS showed an ∼2-fold increased risk of GLP-1 receptor agonist initiation for the top decile versus the middle quintile. The biologically enriched MetPRS has the potential to add an extra layer of information to disease prediction and management approaches for metabolic diseases.