Regulation of PD1 signaling is associated with prognosis in glioblastoma multiforme

Abstract

Glioblastoma is an aggressive cancer of the brain and spine. While analysis of glioblastoma `omics data has somewhat improved our understanding of the disease, it has not led to direct improvement in patient survival. Cancer survival is often characterized by differences in expression of particular genes, but the mechanisms that drive these differences are generally unknown. We therefore set out to model the regulatory mechanisms that associate with glioblastoma survival. We inferred individual patient gene regulatory networks using data from two different expression platforms from The Cancer Genome Atlas (n=522 and 431). We performed a comparative network analysis between patients with long- and short-term survival, correcting for patient age, sex, and neoadjuvant treatment status. We identified seven pathways associated with survival, all of which were involved in immune system signaling. Differential regulation of PD1 signaling was validated in an independent dataset from the German Glioma Network (n=70). We found that transcriptional repression of genes in this pathway—for which treatment options are available—was lost in short-term survivors and that this was independent of mutation burden and only weakly associated with T-cell infiltrate. These results provide a new way to stratify glioblastoma patients that uses network features as biomarkers to predict survival, and identify new potential therapeutic interventions, thus underscoring the value of analyzing gene regulatory networks in individual cancer patients.

gpuZoo: Cost-effective estimation of gene regulatory networks using the Graphics Processing Unit

“…Then, it iteratively leaves out single samples, calculates a model for the population deprived of the i th sample (W (i) ), and uses the difference between these two models to estimate the network for the i th sample ( # ) using equation 4 in (12). In our previous work, we have applied LIONESS to aggregate network models calculated using PANDA (14,17,18). In this case, computing a LIONESS network requires the following steps:…”

“…In our previous work, we have applied LIONESS to aggregate network models calculated using PANDA (14,17,18). In this case, computing a LIONESS network requires the following steps:…”

GRAND: A database of gene regulatory network models across human conditions

“…The colon GRNs were derived using expression data for 12,817 genes from 445 samples in TCGA and 1,193 samples found in GEO as described previously (34) (Figure S2). Glioblastoma networks were generated on 10,701 genes from 953 samples in TCGA and 70 samples across 10,439 genes from the German Glioma Network (GGN) (35). Pancreatic cancer networks were generated from 150 samples from TCGA spanning both basal-like and classical subtypes, across 3,214 genes (36).…”

“…We previously used the netzoo methods, particularly PANDA and LIONESS, to infer tens of thousands of GRN models. We analyzed these networks in a number of published studies, including GRN comparison of 36 "normal" tissues and two cell lines from the Genotype Tissue Expression (GTEx) project (28,32,33) and six cancers from The Cancer Genome Atlas (TCGA) (30,(34)(35)(36). Although each study included detailed descriptions of the data and methods used to generate these networks, there was no appropriate data repository for publishing, querying, and visualizing the GRN models themselves due to the large number of genome-scale networks with millions of edges that required more than 6TB of data storage.…”

Reconstructing Sample-Specific Networks using LIONESS

“…The single-sample networks predicted by LIONESS provide a way to unite (i) the extensive literature and methodologies for estimating complex network relationships using genomic data with (ii) statistical analysis techniques that use sample-level information to model heterogeneity and compare phenotypic groups. Applications of LIONESS have included analyzing the yeast cell cycle (Kuijjer et al, 2019a), studying biological processes in lymphobastoid cell lines (Kuijjer et al, 2019a), investigating the relationship between the host transcriptome and nasal microbiome in infants with respiratory syncytial virus infection (Sonawane et al, 2019), investigating microbiome co-occurrence in respiratory infections (Mac Aogáin et al, 2021), identifying regulatory processes associated with brain cancer survival (Lopes-Ramos et al, 2021) and with sexual dimorphism in colorectal cancer chemotherapy response (Lopes-Ramos et al, 2018), characterizing sex differences in twenty-nine tissues from the Genotype-Tissue Expression Project (Lopes-Ramos et al, 2020), and identifying tissue-specific regulatory processes in maize (Fagny et al, 2020). LIONESS has also been integrated into a method to identify cancer driver genes (Pham et al, 2021).…”