Abstract
Glioblastoma (GBM) is the most common and aggressive malignant brain tumor. Despite treatment advances, the median survival time is still below 2 years since recurrence is nearly universal. Therefore, the discovery of novel and specific molecular targets is needed. However, the identification of such targets is difficult given the high degree of variability among patient samples and individual sequencing analysis is needed for patient specific therapies. Identifying differentially expressed (DE) genes based on small sample sizes is yet another challenge since there is high variability among the different analysis algorithms. To circumvent this problem, we implemented a bioinformatics pipeline, which utilizes the large number of samples in TCGA as a reference to identify DE genes in an individual patient's tumor. By using this algorithm, we produced enriched single patient RNAseq DE gene lists and subsequently calculated a hypergeometric probability and correlation coefficient for every gene pair on these lists. Furthermore, by using the most significant of these pairs, we generated gene association networks. Importantly, our networks were validated utilizing protein-protein interaction studies and therefore could be used to identify patient specific combination therapies for GBM.
