Metabolic Signatures and Diagnostic Prediction Models for Coronary Artery Disease and Type 2 Diabetes Mellitus: Insights From an Exploratory Study

Coronary artery disease (CAD) and type 2 diabetes mellitus (T2DM) often coexist and mutually exacerbate each other through shared mechanisms (including deregulated glucolipid metabolism and insulin resistance), yet their global metabolic signatures remain poorly characterized. In this exploratory study, we collected serum samples from 21 CAD patients, 20 T2DM patients, and 20 healthy controls (HH). Metabolomics and lipidomics were performed, and bioinformatics analysis identified disease‐related pathways and biomarkers. A machine learning‐based diagnostic prediction model was built using these metabolic biomarkers. Our analysis unveiled significant metabolic similarities between CAD and T2DM. Both patient groups had higher covariates such as age and weight compared to the healthy controls. Both conditions also demonstrated elevated levels of triglycerides (TGs), diglycerides (DGs), phosphatidylcholines (PCs), and phosphatidylethanolamines (PEs). Concurrently, lipid metabolism‐related signaling pathways were notably activated. Machine learning identified seven lipid metabolites (e.g., TG 16:0/16:0/18:3, PC 31:1, and PE 34:1) that exhibited an auROC ≥ 0.9 for both CAD and T2DM, suggesting their potential as shared biomarkers. A prediction model based on these seven metabolites achieved predictive accuracies exceeding 95% and 92% for CAD and T2DM, respectively. Furthermore, the model demonstrated predictive accuracies surpassing 98% and 95% in an external validation cohort, highlighting its robust predictive performance. This exploratory study preliminarily delineates the metabolic signatures of CAD and T2DM and strengthens the evidence for their intrinsic relationship from a metabolic perspective.