Abstract Introduction Proteomic data hold promise for predicting future cardiovascular (CV) events. A streamlined proteomics panel with fewer targets could reduce profiling costs and facilitate its integration into clinical practice. Purpose This study aims to identify a subset of proteins that could serve as biomarkers for predicting major adverse cardiovascular events (MACE) in the community. Methods We analysed 768 participants recruited from the general population (mean age 58±11 years, 49.7% women) who underwent targeted proteomic profiling and were followed for MACE over a median of 10.9 years. To identify the most relevant proteins for predicting incident MACE, we applied Random Survival Forests and Gradient Boosting Survival Analysis. Key proteins were selected based on the intersection of the top 25% biomarkers identified by Shapley values from both survival models. Using an unsupervised machine learning approach, participants were clustered based on protein values using a Gaussian Mixture Model. The adjusted hazard ratios of the clusters were assessed through Cox regression analysis. Results Of the 92 available proteins, 15 emerged as key markers in both survival models. These proteins are involved in various biological processes, including blood pressure regulation and kidney function (ACE2, BNP, KIM-1, REN), inflammatory and immune response (CD40L, Dkk1, Gal-9, RAGE), apoptosis (TRAIL-R2), autophagy (CTSL1), proteolysis (PAPPA), angiogenesis (ANGPT1, PGF), extracellular matrix organization (MMP12), and metabolism (AGRP). Participants were categorized into three clusters (Figure 1). After adjustments, clusters 2 and 3 were associated with a significantly increased risk of MACE, with hazard ratios of 1.77 (95% CI: 1.27-2.50, p=0.0008) and 1.61 (95% CI: 1.13-2.28, p=0.0078), respectively (Figure 2). Conclusions We identified 15 proteins that stratified participants into three clinically distinct phenogroups and were effective in identifying individuals at higher risk of developing MACE. These proteins are involved in key CV pathophysiological processes and could be further explored as biomarkers for inclusion in a proteomic panel with fewer targets for MACE prediction.
Santana et al. (Sat,) studied this question.