Abstract

Background Reactive oxygen species (ROS) are critically implicated in ischemic stroke (IS), yet the transcriptional networks and predictive biomarkers underlying ROS dysregulation remain incompletely understood. Methods We integrated two independent microarray cohorts (GSE58294 and GSE16561) to comprehensively analyze ROS-related pathways in IS. Single-sample gene set enrichment analysis (ssGSEA) was used to quantify pathway activity, and weighted gene co-expression network analysis (WGCNA) identified modules associated with ROS dysregulation. Functional enrichment and protein-protein interaction (PPI) network analyses characterized the biological functions of module genes. Elastic Net regression modeling, receiver operating characteristic (ROC) analysis, calibration, and decision curve analysis (DCA) were employed to construct and validate a predictive risk score model. SHapley Additive exPlanations (SHAP) analysis was further applied to interpret gene contributions. Immune cell infiltration was assessed using multiple algorithms, and CLEC4E, the top-ranked gene, was functionally investigated through single-gene GSEA. The OGD/R-treated SH-SY5Y cells and mouse ischemia-reperfusion (I/R) models were established for in vivo and in vitro validation. Results ROS-related pathways were consistently upregulated in IS across both cohorts. WGCNA revealed a robust ROS-associated module (brown module), enriched in immune activation and inflammatory signaling processes. Elastic Net regression identified seven key genes (CLEC4E, SLC8A1, HIST1H4H, BMX, MCEMP1, KREMEN1, ZFP36L2) with strong predictive ability (AUC = 0.81–0.86 across datasets). SHAP analysis highlighted CLEC4E as the most influential contributor, positively associated with IS risk. Immune deconvolution indicated that CLEC4E expression was negatively correlated with B- and T-cell infiltration, while functional analysis linked it to MAPK signaling, RNA degradation, and neutrophil activation pathw