Plasma Proteomic Profiles Predict the Risk of Coronary Artery Disease in the UK Biobank Cohort

Coronary artery disease (CAD) is a complex, multifactorial, and serious condition influenced by genetic, environmental, and lifestyle factors. Thus, it is crucial to develop strategies to predict the risk of CAD for individuals. Plasma proteomics provides a powerful framework for identifying novel biomarkers, discovering potential therapeutic targets, and further improving risk stratification. Here, we examined the association between 2,919 plasma proteins and incident CAD in the UK Biobank cohort (n=35,778). As a result, we identified 576 proteins significantly associated with CAD and found significant alterations in key biological pathways, including signal transduction, immune regulation, and chemotaxis, before CAD onset. Subsequently, we developed machine learning models to predict CAD onset at different time intervals (5 years, 10 years, over 10 years, and entire cohort), demonstrating superior performance over models based on polygenic risk scores (ΔAUC = 0.052), and Pooled Cohort Equations (ΔAUC = 0.049). Notably, the integration of PRS with proteomic data resulted in a marked enhancement in predictive accuracy (AUC = 0.779), comparable to the full model (AUC = 0.780). Key plasma protein predictors, including MMP12, GDF15, and EDA2R, showed sustained importance across models predicting CAD onset at multiple time points. Additionally, Mendelian randomization analysis provided robust evidence for a causal relationship between six plasma proteins and CAD, including MMP12, LPA and PLA2G7, highlighting their potential as therapeutic targets. In conclusion, our study elucidates the plasma proteome associated with CAD, reveals underlying pathogenic mechanisms, and provides valuable insights for identifying high-risk individuals and advancing precision medicine.