Identification of Prognostic Markers of Glioblastoma through Bioinformatics Analysis.

Objective: Glioblastoma is the most common and aggressive type of the central nervous system cancers. Although radiotherapy and chemotherapy are used in the treatment of glioblastoma, survival rates remain unsatisfactory. This study aimed to explore differentially expressed genes (DEGs) based on the survival prognosis of patients with glioblastoma and to establish a model for classifying patients into different risk groups for overall survival.

Methods: DEGs from 160 tumor samples from patients with glioblastoma and 5 nontumor samples from other patients in The Cancer Genome Atlas database were identified. Functional enrichment analysis and a protein-protein interaction network were used to analyze the DEGs. The prognostic DEGs were identified by univariate Cox regression analysis. We split patient data from The Cancer Genome Atlas database into a high-risk group and a low-risk group as the training data set. Least absolute shrinkage and selection operator and multiple Cox regression were used to construct a prognostic risk model, which was validated in a test data set from The Cancer Genome Atlas database and was analyzed using external data sets from the Chinese Glioma Genome Atlas database and the GSE74187 and GSE83300 data sets. Furthermore, we constructed and validated a nomogram to predict survival of patients with glioblastoma.

Results: A total of 3572 prognostic DEGs were identified. Functional analysis indicated that these DEGs were mainly involved in the cell cycle and focal adhesion. Least absolute shrinkage and selection operator regression identified 3 prognostic DEGs (EFEMP2, PTPRN, and POM121L9P), and we constructed a prognostic risk model. The receiver operating characteristic curve analysis showed that the areas under the curve were 0.83 for the training data set and 0.756 for the test data set. The predictive performance of the prognostic risk model was validated in the 3 external data sets. The nomogram showed that the prognostic risk model was reliable and that the accuracy of predicting survival in each patient was high.

Conclusion: The prognostic risk model can effectively classify patients with glioblastoma into high-risk and low-risk groups in terms of overall survival rate, which may help select high-risk patients with glioblastoma for more intensive treatment.