Oxidative stress-driven cellular heterogeneity in Knee Osteoarthritis: scRNA-seq pinpoints SPP1 and CYBA for diagnostic development

Knee osteoarthritis (KOA) is a progressive degenerative joint disease characterized by cartilage degeneration, synovial inflammation, and subchondral bone remodeling. Oxidative stress plays a critical role in KOA pathogenesis. This study aimed to identify oxidative stress-related hub genes by integrating single-cell RNA sequencing (scRNA-seq) with machine learning to uncover potential diagnostic and prognostic biomarkers for KOA. ScRNA-seq data from KOA cartilage tissues were obtained from the GEO database to characterize cellular heterogeneity and identify disease-associated chondrocyte subsets. CellChat was used to analyze intercellular communication, and Monocle2 was applied to reconstruct cell differentiation trajectories. Oxidative stress activity at the single-cell level was evaluated using AUCell, while high-dimensional weighted gene co-expression network analysis (hdWGCNA) was performed to identify oxidative stress-related gene modules. Candidate genes were screened using Venn analysis and further refined by LASSO regression, SVM-RFE, and XGBoost combined with SHAP analysis. Diagnostic performance was assessed, and biological relevance was explored through immune infiltration and pathway analyses. Experimental validation was conducted using qRT-PCR and western blot assays in H2O2-treated chondrocytes. InfC and ProFC were detected specifically in OA cartilage and exhibited high oxidative stress activity. InfC was enriched in inflammatory and immune-related pathways, whereas ProFC was associated with cell cycle regulation and extracellular matrix remodeling. Integrated hdWGCNA and machine-learning analyses consistently identified CYBA and SPP1 as candidate genes. The two-gene model achieved bootstrap-corrected AUCs of 0.937 and 0.825 in the training and validation cohorts, respectively. qRT-PCR and western blot analyses further showed increased CYBA and SPP1 expression in H2O2-treated chondrocytes. Conclusion InfC and ProFC may represent OA-associated chondrocyte states characterized by enhanced oxidative stress, while CYBA and SPP1 may serve as candidate diagnostic biomarkers. Given the small cohort sizes and preliminary in vitro validation, these findings require confirmation in larger independent cohorts and functional studies.