Identification of diagnostic and prognostic hub genes in osteoarthritis: an integrated bioinformatics study

Osteoarthritis (OA) is a heterogeneous whole-joint disease and a leading cause of pain and disability worldwide. Synovial inflammation plays a critical role in OA progression, yet the underlying mechanisms remain unclear and reliable diagnostic biomarkers are lacking.

To identify synovium-based diagnostic biomarkers and explore the molecular mechanisms associated with OA using integrated bioinformatics, machine learning, and experimental validation.

Synovial gene expression profiles from OA and normal controls were integrated and analyzed after normalization and batch-effect correction. Differentially expressed genes (DEGs) were identified, followed by immune infiltration analysis and weighted gene co-expression network analysis. Candidate diagnostic biomarkers were screened using least absolute shrinkage and selection operator, random forest, and support vector machine-recursive feature elimination. Protein-protein interaction network analysis, external dataset validation, and experimental validation in an OA mouse model were further performed.

We identified 574 DEGs and observed marked immune microenvironment remodeling in OA synovium. Integrative analysis yielded 54 candidate genes, from which CX3CR1, KLF9, and MCL1 were identified as candidate diagnostic biomarkers. In the training cohort, these genes showed strong diagnostic performance, with AUCs of 0.983, 0.985, and 0.975, respectively. Network analysis identified 41 interacting genes mainly enriched in inflammation- and tissue remodeling-related pathways. External validation showed that KLF9 had the most robust diagnostic performance.

OA synovium is characterized by substantial immune remodeling and activation of inflammation-related pathways. CX3CR1, KLF9, and MCL1 may serve as candidate diagnostic biomarkers for OA, with KLF9 showing the strongest external validation performance. Key Points • Integrated synovial transcriptomic and bioinformatics analyses identified OA-related differentially expressed genes and co-expression modules. • WGCNA combined with LASSO, SVM-RFE, and RF identified CX3CR1, KLF9, and MCL1 as candidate diagnostic biomarkers for OA. • PPI analysis revealed 41 interacting genes mainly involved in inflammation- and tissue remodeling-related pathways. • Immune infiltration analysis demonstrated substantial synovial immune microenvironment remodeling in OA, supporting a role for immune dysregulation in disease progression.