Construction of a novel microRNA-based signature for predicting the prognosis of glioma

Abstract

Background and purpose: Glioma is a frequent primary brain tumor. MicroRNAs (miRNA) have been shown to potentially play a crucial part in tumor development. Based on miRNAs and clinical factors, a model was constructed to predict the glioma prognosis. Methods: The miRNA expression profiles of glioma come from The Cancer Genome Atlas (TCGA, training group) and Chinese Glioma Genome Atlas (CGGA, validation group). Regression analyses of Cox and Lasso were applied to identity miRNAs associated with glioma prognosis in the TCGA database. The miRNAs were combined with clinical factors to construct individualized prognostic prediction models, whose performance was validated in the CGGA database. The role of miRNA in glioma development was investigated by in vitro experiments. Results: We identified five key miRNAs associated with glioma prognosis and constructed a prediction model. The area under ROC curve for predicting 3-year survival of glioma patients in the TCGA and CGGA groups was 0.844 and 0.770, respectively. The nomogram constructed using the miRNA risk scores and clinical factors showed high accuracy of prediction in the TCGA group (C-index of 0.820) and the CGGA group (C-index of 0.722). The miR-196b-5p altered the migration, proliferation, invasion, and apoptosis of glioma cells by regulating target genes, according to in vitro experiments. Conclusions: A miRNA-based individualized prognostic prediction model was constructed for glioma and miR-196b-5p was identified as a potential biomarker of glioma development.