Abstract
Coronary artery disease (CAD) is a complex condition influenced by genetic factors, lifestyle, and other risk factors that contribute to increased mortality. This study is aimed at evaluating the diagnostic potential of genes associated with cuproptosis, ferroptosis, and pyroptosis (CFP) using network modularization and machine learning methods. CAD-related datasets GSE42148, GSE20680, and GSE20681 were sourced from the GEO database, and genes related to CFP genes were gathered from MsigDB and FerrDb datasets and literature. To identify diagnostic genes linked to these pathways, weighted gene coexpression network analysis (WGCNA) was used to isolate CAD-related modules. The diagnostic accuracy of key genes in these modules was then assessed using LASSO, SVM, and random forest models. Immunity and drug sensitivity correlation analyses were subsequently performed to investigate possible underlying mechanisms. The function of a potential gene, STK17B, was analyzed through western blot and transwell assays. Two CAD-related modules with strong correlations were identified and validated. The SVM model outperformed LASSO and random forest models, demonstrating superior discriminative power (AUC = 0.997 in the blue module and AUC = 1.000 in the turquoise module), with nine key genes identified: CTDSP2, DHRS7, NLRP1, MARCKS, PELI1, RILPL2, JUNB, STK17B, and SLC40A1. Knockdown of STK17B inhibited cell migration and invasion in human umbilical vein endothelial cells. In summary, our findings suggest that CFP genes hold potential as diagnostic biomarkers and therapeutic targets, with STK17B playing a role in CAD progression.