BACKGROUND: Glioblastoma (GBM) is an extremely challenging malignancy to treat. Although temozolomide (TMZ) is a standard treatment regimen, many patients with GBM develop chemoresistance. The aim of this study was to identify a DNA repair-related gene signature to better stratify patients treated with TMZ. METHODS: We selected 89 cases of primary GBM (pGBM) from the Chinese Glioma Genome Atlas RNA-seq dataset as the training cohort, whereas The Cancer Genome Atlas RNA-seq and Gene Set Enrichment (GSE) 16011 mRNA array sets were used as validation cohorts. Regression analysis and linear risk score assessment were performed to build a DNA repair-related signature. We used Kaplan-Meier analysis to evaluate the predictive value of the signature for overall survival (OS) in the different groups. Multivariate Cox regression analysis was used to determine whether the 5-gene signature could independently predict OS. RESULTS: Using our 5-gene signature panel of APEX1, APRT, PARP2, PMS2L2, and POLR2L, we divided patients with pGBM into high-and low-risk groups. Patients with a low-risk score were predicted to have favorable survival and greater benefit from TMZ therapy compared with patients from the high-risk group (P<0.05). Moreover, receiver operating characteristic curves showed that the multigene signature was the most sensitive and specific model for survival prediction (P<0.05). CONCLUSIONS: Among patients with pGBM, classification based on a risk score determined using a 5-gene panel indicated different OS and reaction to TMZ. The findings in this study demonstrate that this unique 5-gene signature could be a novel model to predict OS and provide accurate therapy for patients with pGBM.