Bioinformatics mining of ferroptosis-related biomarkers and development of a diagnostic model in primary Sjögren's syndrome

Objective To identify iron death-related genes in primary Sjögren's syndrome (pSS) through various bioinformatics methods, providing potential targets for the diagnosis and treatment of pSS.

Methods Gene expression matrix of pSS was obtained from the GEO database, and differentially expressed genes (DEGs) were identified using the "limma" package by R software. Weighted gene co-expression network analysis (WGCNA) was used to identify the most relevant modular genes in pSS. Iron death-related gene sets were obtained from the FerrDb database, and key genes were identified by intersecting the most relevant module genes with DEGs and iron death-related genes. Two machine learning algorithms were used to remove redundant genes and construct a diagnostic model consisting of key genes. The accuracy of the diagnostic model was validated in three independent datasets, and the association between key genes and immune cells was revealed by immune infiltration analysis and single-cell data analysis using the Seurat package.

Results A total of 265 DEGs were obtained by differential gene analysis, and the WGCNA algorithm identified a module that was most strongly correlated with pSS (r=0.44, P < 0.01), from which eight iron death-related genes were identified. After removing redundant genes, three key genes related to iron death (PARP9, PARP12, and PARP14) were finally obtained, and a diagnostic model was established based on the three genes. The model exhibited excellent diagnostic performance in three independent datasets (the areas under the receiver operating characteristic curve were 0.848, 0.853 and 1.000, respectively). Immune infiltration analysis and single-cell data analysis revealed significant associations of the three key genes with dendritic cells, macrophages and T/B cells.

Conclusion In this study, three key genes related to iron death in pSS (PARP9, PARP12, PARP14) are identified based on bioinformatics methods and a diagnostic model of pSS is constructed, providing a new potential tool for the diagnosis of pSS.