Introduction: DNA methylation is a mechanism altering the normal state of cells implicated in many cancers. Currently the methylation status of MGMT is one of the most widely utilized clinical genetic tests performed on glioblastoma multiforme (GBM). While several global gene expression signatures have been developed, it is unclear if genome-wide DNA methylation signatures can predict prognosis in cancer.
Methods: We used a computational algorithm (MethylMix) to analyze genome-wide DNA methylation in GBM data obtained from The Cancer Genome Atlas (TCGA). MethylMix identified a set of driver genes that met criteria for being both differential and functional. Differential refers to a difference in cancer methylation compared to normal tissue; functional refers to having a significant correlation with matched gene expression changes. We then used these MethylMix driver genes to build multivariate models of overall survival using linear regression and validated these models in independent data sets.
Results: Applying MethylMix and linear regression we identified a novel methylation signature predictive of overall survival, which we here define as the GLIOMETH signature. Interestingly, GLIOMETH did not include MGMT, suggesting that MGMT methylation is not essential to predict prognosis in GBM. GLIOMETH was prognostically significant even in a multivariate analysis with known prognostic covariates, including MGMT methylation (Figure 1). We validated GLIOMETH in two external DNA methylation data sets and two gene expression data sets, using a leveraging technique predicting methylation in terms of gene expression, showing also a significant survival correlation.
Conclusions: Differential and functional DNA methylation is predictive of overall survival in GBM independent of known prognostic factors. We identified GLIOMETH as a DNA methylation signature that is predictive of overall survival in GBM, outperforming MGMT methylation. The GLIOMETH model validated across multiple independent DNA methylation and gene expression validation data sets demonstrating its robustness as an independent predictor of prognosis in GBM.
Patient Care: GLIOMETH has the potential to improve the prognostic assessment of glioblastoma multiforme based on DNA methylation of diagnostic samples
Learning Objectives: By the conclusion of this session, participants should be able to: 1) Understand the prognostic relationship of genome wide DNA methylation in glioblastoma multiforme. 2) Understand that GLIOMETH is independently prognostic in the presence of known prognostic factors 3) understand that GLIOMETH might be a better alternative to predict survival in GBM compared to MGMT methylation.