Weighted Gene Coexpression Network Analysis Reveals Cancer Stem Cell-Associated Metabolic Gene Signature in Glioma

Abstract Background: Isocitrate dehydrogenase (IDH) mutant glioma patients have a favorable prognosis, accompanying with metabolic alterations and glioma cell dedifferentiation. Recently, mRNA expression-based stemness index (mRNAsi) characteristic relation to IDH status of gliomas has yet illuminated. Thus, we aimed to establish a cancer stem cell-associated metabolic gene signature for risk stratification of gliomas. Methods: The glioma samples came from The Cancer Genome Atlas (TCGA) and the Chinese Glioma Genome Atlas (CGGA) databases. Next, we performed the differential expression analysis between IDH mutant and IDH wild-type gliomas and also conducted weighted gene correlation network analysis (WGCNA) for determining the modules associated with cancer stem cell trait. Subsequently, multivariate Cox regression analysis with the Akaike information criterion (AIC) algorithm was employed to establish a stemness-related metabolic gene signature, which was validated using time-dependent receiver operating characteristic (ROC) curves and concordance index (C-index). Also, we developed a nomogram based on clinical traits and prognostic model. Additionally, according to the results of immunohistochemistry (IHC) staining, the protein levels of gene signature were consistent with the genes expression’s direction.Results: Low expression of mRNAsi was capable of predicting the unfavourable OS of gliomas with a 5-year survival rate of 14.08%. The blue module and its 1466 genes were pertinent to mRNAsi characteristic. Next, Kaplan-Meier (KM) survival curves revealed that cancer stem cell-associated metabolic genes exerted impact on gliomas’ prognosis. Subsequently, univariate and multivariate Cox regression analyses were implemented, and gene signature (LCAT, UST, GALNT13, and SMPD3) was constructed, with C-index of 0.798 (95%CI: 0.769-0.827). Notably, the prognostic model presented a superior predictive value for gliomas’ survival, with the area under the curve (AUC) of ROC curves at 1-year, 3-year as well as 5-year time-point of 0.845, 0.85 and 0.811, respectively. And forest plot uncovered its role as a potential independent predictor for gliomas (HR=2.840, 95%CI: 1.961-4.113, P <0.001). Nomogram also presented superior predictive performance for gliomas’ OS. Conclusion: The gene signature (LCAT, UST, GALNT13, and SMPD3) can be used for risk stratification and also can serve as an independent prognostic factor of glioma patients.