NIMG-46. CHARACTERIZATION OF GLIOBLASTOMA SUBTYPES BY INTEGRATIVE ANALYSIS OF MATCHED GENE EXPRESSION AND ADVANCED MAGNETIC RESONANCE IMAGING DATA

Abstract

The aim of this study was the radiogenomic characterization of glioblastoma subtypes (Verhaak et al. 2010) by integrative analysis of matched magnetic resonance (MR) imaging and gene expression data (GED) in glioblastoma. GED from biopsies obtained by MR-based neuronavigation from 21 glioblastoma patients having had preoperative perfusion, diffusion and MR spectroscopic (MRS) imaging were analyzed using in-house tools, including weighted gene co-expression network analysis (WGCNA) and gene set enrichment analysis (GSEA). Integrative analysis revealed that the proneural glioblastoma subtype tends to have high creatine in MRS, low cerebral blood volume (CVB) and small vessel size (VS) in perfusion imaging, as well as low fractional anisotropy (FA) and high mean diffusivity (MD) in diffusion imaging. The classical subtype showed low values for glutamate and glutamine in MRS, large VS in perfusion as well as high FA and low MD in diffusion imaging. Finally, the mesenchymal subtype had low creatine in MRS, high CBV in perfusion as well as high FA and low MD in diffusion imaging. In addition to the glioblastoma subtype characterization, biologic pathways correlating with different imaging parameters were identified: N-acetylaspartate was linked to oligodendrocytic differentiation and neural development, CBV correlated with genes of the endothelial growth factor and the endothelial-mesenchymal transition (EMT) pathways, VS with hypoxia pathways, FA with genes from the NFkB and the endothelial-mesenchymal transition pathways, and MD was linked to the expression of genes relating to neural development and neural functions. Finally, survival analysis revealed correlations of imaging features with patient prognosis. High N-acetylaspartate, low FA or high MD were linked to a longer progression-free survival in our study population. Although these results remain to be validated in a larger study population our research showed that advanced MR imaging techniques can help discriminate between glioblastoma subtypes and improve prognosis evaluation in glioblastoma.