Prediction of a lncRNA-miRNA co-expression network as a prognostic marker in glioblastoma

Abstract

Abstract Glioblastoma has the characteristics of high proliferation and high metastasis; it has become one of the most difficult tumors to treat in clinic. Recently, more and more attention has been paid to the role of competitive endogenous RNA (cyclin-inhibitor receptor, CeRNA) in cancer. However, there is still limited research on the role of endogenous RNA interactions based on high-throughput sequencing data in the occurrence and development of glioblastoma. In this study, RNA-seq data (miRNA expression profile and lncRNA expression profile) and clinical follow-up data (age, survival time, etc) of human brain low-grade glioblastomas (LGG) were obtained from The Cancer Genome Atlas (TCGA) database. In addition, CIBERSORT was applied to assess immune infiltrates and microenvironmental indicators in LGG. The similarities between the mRNA modules with clinical traits were subjected to weighted correlation network analysis (WGCNA). Based on the mRNAs from clinical-related modules, a survival model was established by univariate and multivariate Cox proportional hazard regression analyses. In the survival model, 519 glioblastoma patients were divided into high-risk (N = 259) and low-risk (N = 260) groups and this model predicts one-year survival in LGG patients, with an AUC of 0.88. Then, we predicted interactions between lncRNAs and miRNAs by starBase. Finally, we identified lncRNA PSMA3-AS1 and miR-143-3p for the construction of a ceRNA network, which might play a key role in the development of glioblastoma and act as a prognostic biomarker of glioblastoma. In this study, we hypothesize that this ceRNA network will impact the immune activity and tumor microenvironment changes in glioblastoma. Therefore, this study provides a new direction for investigating the development and treatment of glioblastoma.