Identification of PSMB9 and CXCL13 as Immune-related Diagnostic Markers for Rheumatoid Arthritis by Machine Learning

Rheumatoid arthritis (RA) is a chronic inflammatory disease that causes significant physical and psychological damage. Although researchers have gained a better understanding of the mechanisms of RA, there are still difficulties in diagnosing and treating RA. We applied a data mining approach based on machine learning algorithms to explore new RA biomarkers and local immune cell status.

We extracted six RA synovial microarray datasets from the GEO database and used bioinformatics to obtain differentially expressed genes (DEGs) and associated functional enrichment pathways. In addition, we identified potential RA diagnostic markers by machine learning strategies and validated their diagnostic ability for early RA and established RA, respectively. Next, CIBERSORT and ssGSEA analyses explored alterations in synovium-infiltrating immune cell subpopulations and immune cell functions in the RA synovium. Moreover, we examined the correlation between biomarkers and immune cells to understand their immune-related molecular mechanisms in the pathogenesis of RA.

We obtained 373 DEGs (232 upregulated and 141 downregulated genes) between RA and healthy controls. Enrichment analysis revealed a robust correlation between RA and immune response. Comprehensive analysis indicated PSMB9, CXCL13, and LRRC15 were possible potential markers. PSMB9 (AUC: 0.908, 95% CI: 0.853-0.954) and CXCL13 (AUC: 0.890, 95% CI: 0.836-0.937) also showed great diagnostic ability in validation dataset. Infiltrations of 16 kinds of the immune cell were changed, with macrophages being the predominant infiltrating cell type. Most proinflammatory pathways in immune cell function were activated in RA. The correlation analysis found the strongest positive correlation between CXCL13 and plasma cells, PSMB9, and macrophage M1.

There is a robust correlation between RA and local immune response. The immune-related CXCL13 and PSMB9 were identified as potential diagnostic markers for RA based on a machine learning approach. Further in-depth exploration of the target genes and associated immune cells can deepen the understanding of RA pathophysiological processes and provide new insights into diagnosing and treating RA.