Identification of purity and prognosis-related gene signature by network analysis and survival analysis in brain lower grade glioma

Abstract

Tumour microenvironment of brain lower grade glioma (LGG) consists of non-tumour cells including stromal cells and immune cells mainly. These non-tumour cells dilute the purity of LGG and play pivotal roles in tumour growth and development, thereby affecting patient prognosis. Tumour purity is also associated with molecular subtypes of LGG. In this study, we discovered the most relevant module to purity by weighted gene co-expression network analysis (WGCNA) and afterwards performed consensus network analysis and survival analysis to filter 61 significant genes related to both purity and prognosis. In turn, we built a simplified model based on the calculation of purity score, and consensus measurement of purity estimation (CPE), with a satisfactory predictive performance by random forest regression. HLA-E, MSN, GNG-5, MYL12A, ITGB4, PDPN, AGTRAP, S100A4, PLSCR1, VAMP5 were selected as the most relevant genes correlating to both purity and prognosis. The risk score model based on the 10 genes could moderately predict patients’ overall survival. These 10 genes, respectively, were positively correlated positively to immunosuppressive cells like macrophage M2, but negatively correlated to patient prognosis, which may explain partially the poor prognosis with low-purity group.