Identification of important biomarkers associated with ferroptosis in diabetic retinopathy via integrative bioinformatics analysis

Abstract

Diabetic retinopathy (DR) is one of the most prevalent and devastating microvascular consequences of diabetes. While the role of ferroptosis in the etiology of DR is acknowledged by some researchers, the precise mechanisms underlying its pathophysiology and the potential therapeutic implications remain to be fully elucidated. The GSE221521 dataset, derived from peripheral blood mononuclear cells (PBMCs), was used to assess differentially expressed genes (DEGs) between controls and DR patients. WGCNA was used to identify gene coexpression modules that are significantly associated with DR. The machine algorithm techniques are used to select genes with significant predictive power. The diagnostic potential of these genes was validated by assessing their performance through ROC curve analysis. The identified key genes were further validated via two external datasets and reverse transcription quantitative polymerase chain reaction (RT‒qPCR). Finally, a total of 3706 DEGs were screened from the GSE221521 dataset. The intersection of 610 genes identified from the 3706 DEGs and WGCNA with 484 ferroptosis genes yielded 16 ferroptosis-related DEGs. These 16 genes were subjected to further analysis via the LASSO and SVM-RFE methods, alongside five additional genes of interest: PLIN2, PIEZO1, HSPA5, HMOX1, and PEBP1. The validation demonstrated that the differences in the expression of PIEZO1 and HSPA5 were statistically significant across both datasets. RT‒qPCR revealed significant upregulation of PIEZO1 and HSPA5 in the peripheral blood of DR patients (P < 0.05). In conclusion, our study identified PIEZO1 and HSPA5 as pivotal genes associated with ferroptosis in the context of DR.

Keywords: Diabetic retinopathy; Ferroptosis; GEO; GSEA; Machine learning; WGCNA.