A large-scale plasma proteomic study reveals the preclinical evolution and potential biomarkers for coronary atherosclerosis

OBJECTIVES Coronary atherosclerosis (CA) can remain subclinical for decades before clinical onset, while the complex pathophysiological mechanisms underlying its progression remain poorly understood. Proteomics offers a novel perspective for elucidating its pathogenic mechanisms and improving risk prediction. METHODS The plasma proteomic data were from 29,020 UK Biobank participants comprising 2,907 Olink-measured proteins. 1,000 iterations of resampling-based univariable Cox regression were performed to identify robust CA-associated proteins. Locally estimated scatterplot smoothing was used to modeled the temporal trajectories of CA-associated proteins. Cox proportional hazards models integrating CA-associated proteins with traditional risk factors were constructed to predict incident CA, with model performance evaluated using the area under the curve (AUC) of receiver operating characteristic (ROC). RESULTS During the 15-year follow-up, 65 proteins were found to be significantly associated with CA risk across the long-term, short-term, and overall follow-up periods. These CA-associated proteins were predominantly involved in collagen-containing extracellular matrix, endoplasmic reticulum lumen, and defense response to bacterium. By integrating multi-stage significance rankings using the robust rank aggregation method, we identified 5 key candidate proteins. Among these proteins, MMP12 remained persistently elevated 15 years before CA diagnosis, whereas GDF15, WFDC2, EDA2R and CST3 exhibited progressive abnormalities approximately 8 to 13 years prior to diagnosis. The model constructed using only the 5 key candidate proteins combined with traditional risk factors achieved high accuracy for 10-year (AUC = 0.749) and 5-year (AUC = 0.722) CA risk prediction. CONCLUSIONS This study provides a temporal proteomic landscape of CA progression and highlights a compact panel of five plasma proteins that effectively predict CA risk. These findings not only advance understanding of the molecular mechanisms driving CA but also offer translational potential for early detection and precision prevention in clinical practice.