Machine learning and single-cell RNA sequencing analyses identify MS-related monocytes and a five-gene candidate biomarker signature.

OBJECTIVE: Multiple sclerosis (MS) is a chronic autoimmune inflammatory disease of the central nervous system (CNS). Based on single-cell RNA sequencing (scRNA-seq) data from experimental autoimmune encephalomyelitis (EAE), this study applied machine learning algorithms combined with integrative bioinformatics methods to identify pivotal biomarkers associated with MS-related monocytes. MATERIALS AND METHODS: Machine learning and scRNA-seq analyses were performed to characterize MS-related monocytes, leading to the identification of five optimally characterized candidate biomarkers associated with pathogenic alterations. The performance of multiple algorithms, such as logistic regression (LogReg), latent Dirichlet allocation (LDA), support vector machine (SVM), Naive Bayes (NB), k-nearest neighbor (KNN), Rpart, and random forest (RF), was evaluated. In addition, the CIBERSORT, single-sample gene set enrichment analysis (ssGSEA), and GSEA algorithms were employed to investigate and define immunological features and biological functions. Finally, quantitative real-time polymerase chain reaction (qRT-PCR) and immunofluorescence were used to validate the expression of the identified genes. RESULTS: Seven machine learning algorithms consistently validated five key genes (COX5A, CTSS, GBP2, IRF7, and PGAM1) as optimally characterized biomarkers. The infiltration profiles of these genes, together with associated immune cell types, provide potential biological underpinnings for the pathogenic alterations observed in MS. CONCLUSION: Collectively, these findings indicate that COX5A, CTSS, GBP2, IRF7, and PGAM1 represent promising biomarkers for MS. The identified gene signature may improve MS diagnosis and risk stratification and provide new insights into monocyte-driven immunopathology.