Metabolomic Differentiation of Tumor Core vs. Edge in Glioma Via Machine Learning

INTRODUCTION: Gliomas exhibit high intra-tumor and inter-patient heterogeneity. Recently, it has been shown that the microenvironment and phenotype differ significantly between glioma core (inner) and edge (invasive) regions. METHODS: Paired glioma core and infiltrating edge samples were obtained from 27 human patients after craniotomy. Liquid-liquid metabolite extraction was performed on the samples, and metabolomic data was obtained via 2DLC-MS/MS. A boosted generalized linear model was employed to predict metabolomic profiles associated with O-6-methylguanine-DNA methyltransferase (MGMT) promoter methylation. RESULTS: A panel of 52 of 160 metabolites were found to significantly differ between the glioma core and edge regions (p = 0.05). Top metabolites with significantly different relative abundances included nicotinamide, creatine, cystathione, D-pantothenic acid, and maleic acid. The machine learning model predicted MGMT promotor methylation status using 5 key metabolic features within the core and edge tissue specimens with AUROCEdge = 0.935 and AUROCCore = 0.959. Top metabolites associated with MGMT status in core included Y-L-glutamyl-L-glutamic acid, D-pantothenic acid, uric acid, and in edge included itaconic acid, N,N-dimethylarginine, cytosine. CONCLUSIONS: Key metabolic differences are identified between core and edge tissue in glioma and can predict MGMT status.