CD48, CD69, and TIGIT as diagnostic biomarkers for primary Sjögren's syndrome: an integrated machine learning and...

Primary Sjögren's syndrome (pSS) is a chronic systemic autoimmune disorder. However, current diagnostic methods remain limited, necessitating the exploration of non-invasive diagnostic markers with higher specificity.

This study integrated two GEO expression datasets to identify differentially expressed genes (DEGs) specific to pSS (distinct from SLE) and applied LASSO, XGBoost, RF, and SVM-RFE algorithms to screen candidate genes. Correlation and interaction network analyses were performed, followed by construction and validation of a diagnostic nomogram. The model's differential diagnostic ability was validated in IgG4-RD, RA, SLE, and SSc cohorts. Additionally, candidate genes and the diagnostic model were experimentally validated using RT-qPCR in clinical samples.

Three candidate genes (CD48, CD69, and TIGIT) were identified, showing significant upregulation in pSS (individual AUC > 0.80). The combined diagnostic model achieved an AUC of 0.924, with AUC > 0.90 in validation sets, efficiently distinguishing pSS from IgG4-RD, RA, SLE, and SSc. RT-qPCR confirmed their high expression in pSS, with the model yielding AUC 0.875 (accuracy/precision > 0.85). Notably, combining these candidate genes with erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) yielded an AUC of 0.876 and a specificity of 83.3%, outperforming conventional markers such as ANA, anti-SSA, and anti-SSB antibodies.

CD48, CD69, and TIGIT were identified as potential diagnostic markers for pSS. The combined model significantly enhanced diagnostic accuracy and effectively differentiated pSS from other autoimmune conditions. Integration with ESR/CRP substantially improved specificity compared to conventional serological markers.