Construction of a PCR-based panel of gene expression and DNA methylation features for the classification of Glioblastoma subtypes

Abstract

[eng] Glioblastoma (GBM) is the most prevalent and aggressive primary brain tumor, with a median survival of around 15 months. Advancements on multi-omic profiling in combination with computational algorithms have unraveled the existence of four GBM molecular subtypes (Classical, Mesenchymal, Neural, and Proneural). However, GBM patients are not currently classified in clinical settings due to the high cost of the current classification techniques. Using supervised machine learnings with publicly available gene expression and DNA methylation data from The Cancer Genome Atlas (TCGA), we have constructed robust classifiers that require a minimum number of genomic features. Thus, to improve the translation of these findings and increase the chances for a nearterm application in our health system, we developed PCR-based panels that survey the informative gene expression or DNA methylation markers. This approach provides a novel cost-effective method to stratify GBM patients for precision medicine. These panels have the potential to be used in a daily clinical setting to classify GBM patients and potentially adapt the therapeutic management. As a corollary of our study, we found that each GBM subtype presents unique DNA methylation patterns and pathways activation. These findings posed us to explore alternative, GBM subtype-specific drugs that target these pathways, thus potentially increasing the efficacy of currently used drugs and extending the survival of these patients.