Prognostic modeling of glioma using epilepsy-related genes highlights PAX3 as a regulator of migration and vorinostat sensitivity

This study aimed to construct and validate a prognostic model for glioma based on epilepsy-related genes (ERGs) and to investigate the functional role of PAX3 in glioma progression and drug response. Transcriptomic and clinical data from TCGA, GEO, and CGGA databases were used to identify differentially expressed ERGs between glioma patients with and without epilepsy. Univariate Cox regression, LASSO regression, and multivariate Cox analysis were employed to establish a four-gene prognostic model comprising PAX3, RETN, VEPH1, and HTR1A. Patients were stratified into high- and low-risk groups based on the median risk score, which was calculated using gene expression levels and corresponding regression coefficients. The model showed robust prognostic performance, with AUC values exceeding 0.85 in the training set and remaining above 0.73 in internal and external validation cohorts. Kaplan–Meier survival analysis demonstrated significantly longer overall survival in the low-risk group. The risk score was also validated as an independent prognostic factor across multiple datasets. A nomogram integrating clinical features and risk score further improved prediction accuracy, with C-index values up to 0.843 and high calibration concordance. Among the ERGs, PAX3 showed the strongest correlation with the risk score and was overexpressed in glioma, where it promoted proliferation, migration, epithelial–mesenchymal transition, and resistance to vorinostat through regulation of HDAC1/2/3 targets, as confirmed by functional assays showing that PAX3 knockdown suppressed proliferation and migration, while overexpression enhanced these effects. In conclusion, this study developed and validated a four-gene ERG-based prognostic model with high clinical utility and identified PAX3 as a potential therapeutic target that drives glioma cell migration and vorinostat sensitivity.