ABSTRACT: Coronary atherosclerosis (CA) is a leading cause of cardiovascular diseases with the high morbidity and mortality; however, the current diagnostic methods, primarily based on symptoms, signs, lab examination and imaging, are often inadequate for detecting subclinical or early-stage CA, costly, and inaccessible in many cases. The objective of this study was to discover sensitive and specific biomarkers for the diagnosis of CA severity. We enrolled 443 participants, including CA patients and healthy controls, from three independent cohorts: discovery, testing, and blinded validation. Multi-omics data integration during the discovery phase identified key features of atherosclerotic progression and potential biomarkers. Biomarker panels were refined using random forest models in the testing cohort, and their performance was evaluated in a blinded validation cohort to assess their ability to monitor the occurrence and development of CA. Multi-omics analysis revealed that plasma metabolites exhibited the strongest correlation with CA severity, effectively distinguished different CA stages from healthy controls. Post hoc analysis confirmed the diagnostic model's robustness, with an AUC value higher than .933 (95% CI: .828-.984, sensitivity 93.75%, and specificity 80%). In the blinded validation cohort, the biomarker panel achieved AUC values of .821-.898 for CA occurrence and .649-.849 for CA severity. Notably, 90% of these biomarkers remained significant after adjusting for comorbidities (p < .05). This study identified significant metabolic changes during CA progression and established biomarker panels with potential diagnostic value for assessing CA severity. Key metabolites including cholesteryl sulphate, azelaic acid, tryptophan, arabinofuranosyluracil, TMAO, ADMA, LPC18:2, tartaric acid, L-citrulline, and L-proline, purine, sorbitol, and 2-aminoadipic acid. These findings highlight the potential of these biomarkers to improve early diagnosis and personalised management of CA.
