Identification of BMX as a Potential Biomarker for Rheumatoid Arthritis Based on WGCNA, Machine Learning, and Experimental Validation

BACKGROUND: Rheumatoid arthritis (RA) is a chronic autoimmune disease needing reliable biomarkers. This study aimed to identify potential RA biomarkers using WGCNA and machine learning, and analyze their correlation with disease activity and STAT3. METHODS: GEO datasets were used. WGCNA and three machine learning methods (LASSO, SVM, Boruta) screened common core genes. Correlation with STAT3 was analyzed. RT-qPCR validated core gene expression in RA patients. Correlation with clinical parameters (DAS28, CCP, RF, ESR) and ROC curves were assessed. RESULTS: WGCNA identified 11 modules; the black module correlated most with RA. LASSO, SVM, and Boruta identified 16, 35, and 46 key genes respectively, with one overlapping core gene: BMX. BMX expression positively correlated with STAT3. RT-qPCR confirmed BMX upregulation in RA, which positively correlated with DAS28 and RF. ROC analysis gave an AUC of 0.789. CONCLUSION: BMX expression is upregulated in RA patients, correlates with disease activity, and represents a potential diagnostic biomarker. The BMX‑STAT3 correlation suggests its possible involvement in RA pathogenesis, but the regulatory relationship requires further functional validation.