Identification and experimental validation of glycosylation related differentially expressed genes as diagnostic biomarkers in systemic lupus erythematosus.

As a heterogeneous autoimmune disorder, systemic lupus erythematosus (SLE) involves poorly characterized etiological mechanisms and disease pathways. While glycosylation's impact on immune homeostasis and disease pathogenesis has become a focal point in contemporary research, delineating the specific contribution of associated genes to SLE requires expanded investigation. This study uses bioinformatics methods to explore the potential diagnostic value of glycosylation-related differentially expressed genes (GRDEGs) in SLE and verify their differential expression in peripheral blood between patients and healthy individuals through RT-qPCR. Data were obtained from several GEO datasets, specifically GSE50772, GSE81622, and GSE20864. Through expression differential analysis, a total of 26 GRDEGs were detected. The diagnostic utility of these genes was assessed using receiver operating characteristic (ROC) curves, while enrichment evaluations for Gene Ontology and KEGG pathways were conducted to elucidate their functional properties. Additionally, a predictive model for SLE that integrated 14 GRDEGs was developed using logistic regression, LASSO regression, and support vector machines (SVM), and its performance was evaluated on both the training dataset and an external validation cohort. Finally, RT-qPCR was utilized to verify the expression levels of RNASE2, PTGDS, CXCL2, TNFRSF21, and LAMP3 in peripheral blood mononuclear cells collected from SLE subjects and normal donors. Employing differential expression profiling, 26 GRDEGs were identified. Subsequently, their capacity for distinguishing SLE from controls was demonstrated using ROC curves, with discriminative power reflected in AUC scores between 0.7 and 0.9. Functional assessments of these GRDEGs using Gene Ontology and KEGG pathway annotation chiefly indicated association with immune functions, modulation of phagocytic processes, and inflammatory cascades, including but not limited to interleukin-17 and tumor necrosis factor signal transduction pathways. The SLE diagnostic model, constructed using logistic regression, SVM, and LASSO regression, incorporates 14 GRDEGs and demonstrates high robustness in both the training dataset and external validation. Furthermore, RT-qPCR results showed significant expression differences of RNASE2, PTGDS, CXCL2, TNFRSF21, and LAMP3 in SLE patients compared to healthy individual. This study identifies glycosylation-related gene candidates and establishes a preliminary diagnostic model for SLE, utilizing integrated bioinformatics analysis and experimental validation. These findings provide foundational insights for further exploration of SLE's molecular mechanisms and diagnostic advancement.