An Integrated Multi-Omics Analysis Identifies Oxeiptosis-Related Biomarkers in Diabetic Retinopathy

Background: Diabetic retinopathy (DR), a leading cause of blindness, lacks early biomarkers and mechanism-targeted therapies. While oxidative stress drives DR pathogenesis, the role of oxeiptosis—a reactive oxygen species-induced, caspase-independent cell death pathway—remains largely unexplored. Methods: We integrated transcriptomic profiling (GSE221521: 69 DR vs. 50 controls), two-sample Mendelian randomization (MR) using blood cis-eQTLs (GTEx) as instruments and DR GWAS (FinnGen R12) as outcome, machine learning-based feature selection (SVM-RFE and Boruta algorithms), and single-cell RNA sequencing (scRNA-seq) analysis (GSE165784). Functional enrichment, immune deconvolution (CIBERSORT), and diagnostic nomogram construction were performed. We validated the key genes using human retinal microvascular endothelial cells (hRMECs) treated with high glucose (30 mM). Results: Oxeiptosis scores were elevated in DR blood samples (p < 0.001). MR analysis identified five putative causal genes: CASP2 (OR = 1.067), PLEC (OR = 1.035) and FBN2 (OR = 1.016) as risk factors, and CYP27A1 (OR = 0.960) and GPD2 (OR = 0.958) as protective factors. SVM-RFE and Boruta algorithms confirmed CASP2 and PLEC as hub genes. A nomogram incorporating both genes achieved robust DR prediction (AUC = 0.811). Functional analysis associated these genes with innate immune activation and extracellular matrix reorganization. Single-cell transcriptomics revealed PLEC was markedly overexpressed in disease-relevant cells (fibroblasts, endothelial cells), whereas CASP2 exhibited a distinct pattern, with notable enrichment in retinal CD8+ T cells. Both genes were associated with a pro-inflammatory shift in the immune landscape. Their upregulation was validated in independent datasets and high-glucose-stressed retinal cells. Conclusions: This study establishes an integrated multi-omics framework implicating oxeiptosis-related pathways in DR and nominates CASP2 and PLEC as putatively causal, biologically relevant candidate biomarkers and potential therapeutic targets.