To optimize treatment strategies for patients with glioblastoma, a more precise understanding of the molecular basis of this disease clearly is necessary. Therefore, numerous studies have focused on the molecular biology of glioblastoma and its linkage to clinical behavior. Here we investigated 70 glioblastomas using the array-based comparative genomic hybridization (array-CGH) with GenoSensor Array 300 to identify recurrent DNA copy number imbalances associated with patient outcomes. Univariate log-rank analysis of array-CGH data revealed 46 copy number aberrations (CNAs) associated with outcome. Among them, 26 CNAs were associated with shortened survival whereas the remaining 20 CNAs correlated with good prognosis. A hierarchical cluster analysis disclosed two genetically distinct groups of glioblastomas (1 and 2; 56 and 14 tumors, respectively). Univariate log-rank test discerned significant difference in survival between both genetic subsets while the 5-year survival rate consisted of 0 for group 1 and 63% for group 2. Multivariate analysis revealed that unfavorable genetic signature is an independent prognostic factor increasing a risk of patient death (hazard ratio, 4.38; P=0.00001). In conclusion, our current study suggests that glioblastomas can be subdivided into clinically relevant genetic subsets. Therefore, array-CGH screening of glioblastomas could provide clinically useful information and, perhaps, potentially improve the quality of treatment.