Abstract
Glioblastoma is the most common and aggressive primary brain tumor in adults, with poor survival despite standard therapy. Current risk stratification methods have been largely unsuccessful, and no improvements to the standard of care have emerged in over two decades. To address this, we developed a transcriptome-based model to identify patients at high risk of early progression and to inform treatment decisions. RNA-sequencing data from tumor tissue of 75 patients (median age 61 years, 44 males) with newly diagnosed glioblastoma were analyzed alongside progression-free survival (PFS) and overall survival (OS) outcomes. A bootstrap-based Cox regression was used to identify genes consistently associated with survival, followed by regularized modeling to derive a transcriptomic risk score. Patients were stratified into high- and low-risk groups using the median risk score. Multivariable Cox regression, adjusting for age, MGMT status, and gender, confirmed the risk score as an independent prognostic factor. The risk score was significantly associated with worse outcomes, with a hazard ratio of 2.93 for PFS (p = 4.9 × 10⁻⁸; concordance = 0.767) and 1.89 for OS (p = 0.0029; concordance = 0.751). Pearson correlation revealed strong negative associations between the risk score and both PFS (r = –0.76, p = 3.4 × 10⁻¹⁵) and OS (r = –0.52, p = 2.0 × 10⁻⁶). Classification based on the risk score achieved 79% sensitivity and 91% specificity for identifying early progression, and 63% sensitivity and 65% specificity for OS. Notably, the model includes genes such as MGMT (HR = 2.07), GINS4 (HR = 2.05), and GPNMB (HR = 1.43), which are involved in DNA repair, cell cycle regulation, and tumor invasion. Downregulation of MMP10 (HR = 0.36) was associated with improved prognosis. These findings support the clinical utility of this transcriptome-derived model in guiding personalized treatment strategies for glioblastoma.