Abstract
BACKGROUND: Coronary angiography remains the gold standard for diagnosing coronary artery disease (CAD), but its invasive nature limits its applicability for widespread screening. Identifying non-invasive molecular markers could improve CAD classification and risk assessment. METHODS: We employed 5hmC-Seal technology to generate genome-wide 5-hydroxymethylcytosine (5hmC) profiles from plasma cell-free DNA (cfDNA) in 724 CAD patients, stratified into stable CAD (sCAD), non-ST-elevation myocardial infarction (NSTEMI), and ST-elevation myocardial infarction (STEMI) groups. Using machine learning algorithms, we identified inflammation-related 5hmC modifications associated with disease severity and constructed a classification model based on key hydroxymethylated markers. The model was validated internally and externally in an independent cohort of 167 patients. RESULTS: We found that inflammation-related differentially hydroxymethylated genes (DhMGs) were significantly associated with CAD severity, with enriched pathways linked to immune activation and inflammatory response regulation. A 19-marker panel of 5hmC-based features effectively distinguished CAD patients at different disease stages, with high classification accuracy (AUC = 0.913 in the internal validation cohort). External validation confirmed the robustness of the model, achieving AUCs of 0.784, 0.880, and 0.918 when differentiating between NCA vs. sCAD, sCAD vs. MI, and NCA vs. MI, respectively. Compared to traditional clinical indicators, the 5hmC-based model demonstrated superior discriminatory power for MI. CONCLUSIONS: Our findings suggest that 5hmC modifications in cfDNA reflect CAD-related epigenetic changes and may serve as a promising biomarker for disease stratification. These results provide new insights into the epigenetic landscape of CAD and highlight the potential utility of 5hmC profiling for non-invasive disease monitoring. Further studies are warranted to validate these findings and explore their clinical implications.