Multi-omics integration predicts the incidence of 17 diseases in the UK Biobank

Abstract

Multi-omics technologies, such as metabolomics and proteomics, offer deep molecular perspectives that could enhance risk prediction, but large-scale studies integrating both are scarce. Here we show the predictive values of these two omics across 17 incident diseases in 23,776 UK Biobank participants with complete baseline for 159 NMR-based metabolites and 2,923 Olink affinity-based proteins. We found that adding omics data significantly improved risk prediction for all 17 diseases compared to clinical predictors alone. Proteomics-only models generally outperformed metabolomics-only models for 16 of the 17 diseases, and integrating both omics added little prediction power over proteomics-only models. Furthermore, we identified key omics features, including both well-established (e.g., KLK3/PSA for prostate cancer) and potential novel ones (e.g., PRG3 for skin cancer). We further connected diseases with medication and socioeconomic factors through key proteins, highlighting the clinical utility of omics data for enhancing individual risk prediction, providing molecular insights into disease mechanisms, and potentially guiding future therapeutic development.