Background: Osteoarthritis (OA) is a progressive joint disease influenced by macrophage-associated efferocytosis. Objective: This study aimed to identify key OA biomarkers by integrating transcriptomic and single-cell RNA sequencing (scRNA-seq) data with Mendelian randomization (MR). Materials and Methods: We integrated three OA-related transcriptomic datasets (GSE89408, GSE55457, and GSE152805) from synovial tissue with efferocytosis- and macrophage polarization-related genes (ERGs and MPRGs). Differentially expressed genes (DEGs) were identified and analyzed using gene set enrichment (ssGSEA), weighted gene co-expression network analysis (WGCNA), and functional enrichment (GO/KEGG). MR was performed with GWAS data to evaluate causal links between candidate genes and OA. Machine learning methods (LASSO and SVM-RFE), supported by ROC analysis and artificial neural network (ANN) modeling, and were used to screen potential biomarkers. Single-cell RNA-seq analysis was conducted with Seurat, incorporating clustering, functional enrichment, cell-cell communication (CellChat), and pseudotime trajectory analysis (Monocle) to characterize cellular heterogeneity and differentiation pathways in OA. Results: MR and machine learning highlighted FMO4 and GPR65 as risk biomarkers, validated in independent datasets (AUC > 0.7). GPR65 showed strong associations with neutrophils, regulatory T cells, and antigen-presenting cell (APC) co-inhibition, while both genes were enriched in immune and metabolic pathways. Single-cell analysis identified synovial subintimal fibroblasts and HLA-DRA+ cells as key contributors with distinct differentiation patterns. An ANN model further improved OA classification. Conclusion: These findings provide new insights into OA pathogenesis and support FMO4 and GPR65 as promising diagnostic and therapeutic targets.
Zhang et al. (Wed,) studied this question.