Construction of an Osteoarthritis Diagnostic Model Based on Hub Immune Cells and Genes by Machine Learning Method

The objective of this investigation was to develop a diagnostic model for osteoarthritis (OA) by integrating immune cell profiling with transcriptomic signatures. Four gene expression datasets related to OA were downloaded from the GEO database. CIBERSORT was employed to evaluate the proportion of different immune cell types. Hub immune cells were selected using three distinct optimization algorithms (LASSO, RFE, and RF). Differentially expressed genes (DEGs) between OA and control samples were screened using the limma package. Subsequently, function analysis and protein-protein interaction (PPI) analysis were conducted. Topology analysis based on four algorithms was performed, and hub genes were identified by overlapping the results of these four algorithms. A diagnostic model was constructed and validated using the ROC curve method. Pearson correlation coefficients between hub immune cell populations and candidate genes were computed using the cor() function in R. Seven types of differentially abundant immune cells were identified between the two groups. After analysis with the RF, LASSO, and RFE algorithms, five overlapping immune cells, namely, T cell CD4 memory resting, NK cell activated, T cell CD4 naive, mast cell resting, and mast cell activated, were selected as hub immune cells. A total of 578 DEGs were selected, which were implicated in the MAPK signaling pathway, focal adhesion, and osteoclast differentiation. Following PPI analysis, five hub genes (CXCL8, EEF1A1, IL1B, EEF2, and IL6) were obtained. The diagnostic model demonstrated excellent performance. Significant correlations were observed between the hub genes and immune cell populations. Through systematic analysis, we identified five key immune cell types and five hub genes associated with immune infiltration in OA. These biomarkers were subsequently utilized to construct a diagnostic prediction model.