Transcriptomic Features Predict Survival in Glioblastoma Patients.

backgroundGlioblastoma (GBM) is the most aggressive primary malignant brain tumor in adults and demonstrates substantial transcriptomic heterogeneity associated with patient survival. High-dimensional genomic datasets present statistical challenges because the number of measured molecular features often exceeds the number of available patient samples. Penalized regression methods such as Least Absolute Shrinkage and Selection Operator (LASSO) regression enable simultaneous feature selection and survival modeling in these settings.

methodsTranscriptomic expression and survival data from The Cancer Genome Atlas (TCGA) GBM cohort were analyzed using LASSO-regularized Cox proportional hazards regression implemented through the glmnet package in R. After preprocessing and sample matching, 518 tumor samples and 12,042 transcriptomic features were retained for analysis. Patients were randomly divided into training (n=414) and test (n=104) cohorts. Ten-fold cross-validation using Harrell's concordance index (C-index) was performed to identify optimal regularization parameters. Kaplan-Meier survival analysis was used to evaluate risk stratification performance, and receiver operating characteristic (ROC) analysis was performed for selected candidate genes.

resultsCross-validation identified a maximum C-index of 0.583 at the optimal lambda.min regularization parameter. The final LASSO-Cox model retained 74 genes with nonzero coefficients. Two identified genes, CLEC5A and RANBP17, overlapped with previously reported GBM prognostic genes from an independent study. Kaplan-Meier analysis demonstrated significant survival separation between predicted high-risk and low-risk groups in both the training cohort (log-rank p < 0.0001) and the held-out test cohort (log-rank p < 0.05). ROC analysis of selected candidate genes demonstrated moderate discriminatory performance, with area under the curve values ranging from 0.604 to 0.701.

conclusionsLASSO-regularized Cox regression identified sparse transcriptomic features associated with survival in TCGA GBM patients. Survival stratification performance persisted in a held-out test dataset, supporting the reproducibility of the derived transcriptomic risk model. These findings demonstrate the applicability of penalized survival modeling approaches to high-dimensional cancer transcriptomic datasets and support further investigation of transcriptomic biomarkers in GBM prognosis.