Abstract: BACKGROUND: Mitochondrial dysfunction and ferroptosis are widely involved in the development of atherosclerosis. Research on biomarkers related to mitochondrial dysfunction and ferroptosis in atherosclerosis is important for disease diagnosis and treatment. 
OBJECTIVE: To investigate mitochondrial dysfunction- and ferroptosis-related biomarkers in the pathogenesis of atherosclerosis using bioinformatics and machine learning algorithms, and to predict potential regulatory traditional Chinese medicines (TCMs). 
METHODS: The atherosclerosis dataset GSE100927 was obtained from the GEO database (which is the gene expression database, developed by the National Center for Biotechnology Information in 2000, collects and organizes gene expression data submitted by research institutions and scientists worldwide), and differentially expressed genes were identified, followed by immune infiltration analysis. Weighted correlation network analysis (WGCNA) identified atherosclerosis-related module genes. These module genes were intersected with mitochondrial dysfunction genes, ferroptosis genes, and differentially expressed genes. Consensus clustering analysis was performed on the disease group data based on the intersected genes, and differentially expressed genes between the clusters were identified. Enrichment analysis of the differentially expressed genes was conducted. Hub genes were screened using machine learning algorithms, including the least absolute shrinkage and selection operator (LASSO) and random forest. An atherosclerosis cell model was constructed using RAW264.7 cells, and hub genes were validated through qPCR. Finally, databases were used to predict TCMs regulating the Hub genes.
RESULTS AND CONCLUSION: Five intersected genes were identified by intersecting differentially expressed genes, WGCNA module genes, and mitochondrial dysfunction- and ferroptosis-related genes. Consensus clustering analysis based on these five genes identified two subtypes. Differential analysis between the subtypes revealed 994 subtype-specific differentially expressed genes related to mitochondrial dysfunction and ferroptosis. Three hub genes, dematin actin binding protein (DMTN), Fc gamma receptor IIIa (FCGR3A), and microsomal glutathione S-transferase 1 (MGST1), were predicted using two machine learning algorithms. Experimental validation suggested that DMTN and MGST1 have significant diagnostic values. TCM prediction results indicated that Angelica sinensis and cinnamon may regulate DMTN and MGST1. To conclude, DMTN and MGST1 have diagnostic values and can serve as characteristic genes related to mitochondrial dysfunction and ferroptosis in atherosclerosis. TCMs such as Angelica sinensis and cinnamon have potential as regulators of these genes and warrant further research.

Key words:  bioinformatics, machine learning, atherosclerosis, mitochondrial dysfunction, ferroptosis, immune cells, traditional Chinese medicines