The Cancer Genome Atlas Glioblastoma Research Articles

IDENTIFICATION AND PRELIMINARY TREATMENT DATA OF XENOGRAFTS REPRESENTING TCGA-DEFINED SUBTYPES

BACKGROUND: The Cancer Genome Atlas (TCGA) Network recently catalogued recurrent genomic abnormalities in glioblastoma (GBM). This genomic profiling lead to the molecular classification of GBMs into four subtypes: Proneural, Neural, Classical and Mesenchymal. The importance of identifying these subtypes stands to help researchers and clinicians to better understand GBMs with the potential to personalize treatment options and explore different therapeutic approaches that each subtype may require. As such, the development of preclinical xenograft models that represent each TCGA subtype serves to expedite the evaluation of targeted therapeutic strategies for the treatment of GBMs. METHODS: For this project, established xenografts from patient derived tissue were passaged until reliable growth characteristics were obtained. Established GBM xenografts were classified into TCGA-defined subtypes first by obtaining global gene expression data using Affymetrix. Microarray data was then normalized and probes were summarized as gene expression levels using RMA. Then data was log2 transformed and genes were median centered. Xenografts were subsequently classified into one of four previously defined subtypes as described using Classification to the Nearest Centroid (ClaNC) with the TCGA GBM training dataset. Using established xenografts from each subtype, a preliminary panel of treatment agents was assessed by delay in tumor growth and by tumor regression in athymic mice bearing subcutaneous tumors. Statistical analysis was performed using a personalized SAS statistical analysis program, the Wilcoxon rank order test for growth delay, and Fisher's exact test for tumor regression. RESULTS: To date, our lab has established and identified 2 patient derived xenografts for each TCGA-defined subtype. These xenografts lines have reliable growth patterns with a tumor take rate >80%. Phase I drug testing has been completed and includes tumor growth responses to the following single agents: temozolomide, bevacizumab, BCNU, cytoxan, and irinotecan. CONCLUSIONS: A panel of patient derived xenografts with reliable growth characteristics that represent the four TCGA-defined GBM subtypes has been established and utilized for drug evaluation. The identification of these valid xenograft models represents an important contribution for modeling and predicting therapeutic response. We have currently completed the first round of drug testing within each subtype and look to expand testing to include additional agents as well as select combinations. SECONDARY CATEGORY: n/a.

Integrative network-based Bayesian analysis of diverse genomics data.

BackgroundIn order to better understand cancer as a complex disease with multiple genetic and epigenetic factors, it is vital to model the fundamental biological relationships among these alterations as well as their relationships with important clinical outcomes.MethodsWe develop an integrative network-based Bayesian analysis (iNET) approach that allows us to jointly analyze multi-platform high-dimensional genomic data in a computationally efficient manner. The iNET approach is formulated as an objective Bayesian model selection problem for Gaussian graphical models to model joint dependencies among platform-specific features using known biological mechanisms. Using both simulated datasets and a glioblastoma (GBM) study from The Cancer Genome Atlas (TCGA), we illustrate the iNET approach via integrating three data types, microRNA, gene expression (mRNA), and patient survival time.ResultsWe show that the iNET approach has greater power in identifying cancer-related microRNAs than non-integrative approaches based on realistic simulated datasets. In the TCGA GBM study, we found many mRNA-microRNA pairs and microRNAs that are associated with patient survival time, with some of these associations identified in previous studies.ConclusionsThe iNET discovers relationships consistent with the underlying biological mechanisms among these variables, as well as identifying important biomarkers that are potentially relevant to patient survival. In addition, we identified some microRNAs that can potentially affect patient survival which are missed by non-integrative approaches.

MR imaging predictors of molecular profile and survival: multi-institutional study of the TCGA glioblastoma data set.

To conduct a comprehensive analysis of radiologist-made assessments of glioblastoma (GBM) tumor size and composition by using a community-developed controlled terminology of magnetic resonance (MR) imaging visual features as they relate to genetic alterations, gene expression class, and patient survival. Because all study patients had been previously deidentified by the Cancer Genome Atlas (TCGA), a publicly available data set that contains no linkage to patient identifiers and that is HIPAA compliant, no institutional review board approval was required. Presurgical MR images of 75 patients with GBM with genetic data in the TCGA portal were rated by three neuroradiologists for size, location, and tumor morphology by using a standardized feature set. Interrater agreements were analyzed by using the Krippendorff α statistic and intraclass correlation coefficient. Associations between survival, tumor size, and morphology were determined by using multivariate Cox regression models; associations between imaging features and genomics were studied by using the Fisher exact test. Interrater analysis showed significant agreement in terms of contrast material enhancement, nonenhancement, necrosis, edema, and size variables. Contrast-enhanced tumor volume and longest axis length of tumor were strongly associated with poor survival (respectively, hazard ratio: 8.84, P = .0253, and hazard ratio: 1.02, P = .00973), even after adjusting for Karnofsky performance score (P = .0208). Proneural class GBM had significantly lower levels of contrast enhancement (P = .02) than other subtypes, while mesenchymal GBM showed lower levels of nonenhanced tumor (P < .01). This analysis demonstrates a method for consistent image feature annotation capable of reproducibly characterizing brain tumors; this study shows that radiologists' estimations of macroscopic imaging features can be combined with genetic alterations and gene expression subtypes to provide deeper insight to the underlying biologic properties of GBM subsets.

Integrative Subtype Discovery in Glioblastoma Using iCluster

Large-scale cancer genome projects, such as the Cancer Genome Atlas (TCGA) project, are comprehensive molecular characterization efforts to accelerate our understanding of cancer biology and the discovery of new therapeutic targets. The accumulating wealth of multidimensional data provides a new paradigm for important research problems including cancer subtype discovery. The current standard approach relies on separate clustering analyses followed by manual integration. Results can be highly data type dependent, restricting the ability to discover new insights from multidimensional data. In this study, we present an integrative subtype analysis of the TCGA glioblastoma (GBM) data set. Our analysis revealed new insights through integrated subtype characterization. We found three distinct integrated tumor subtypes. Subtype 1 lacks the classical GBM events of chr 7 gain and chr 10 loss. This subclass is enriched for the G-CIMP phenotype and shows hypermethylation of genes involved in brain development and neuronal differentiation. The tumors in this subclass display a Proneural expression profile. Subtype 2 is characterized by a near complete association with EGFR amplification, overrepresentation of promoter methylation of homeobox and G-protein signaling genes, and a Classical expression profile. Subtype 3 is characterized by NF1 and PTEN alterations and exhibits a Mesenchymal-like expression profile. The data analysis workflow we propose provides a unified and computationally scalable framework to harness the full potential of large-scale integrated cancer genomic data for integrative subtype discovery.

Abstract 3968: Exploring cancer datasets in the integrative genomics viewer (IGV)

Abstract Cancer genome characterization studies, such as The Cancer Genome Atlas (TCGA) and the International Cancer Genome Consortium (ICGC), are producing a flood of diverse data including whole-genome sequencing scans, expression profiles, high-resolution SNP and copy number data, and epigenetic profiles. These diverse datasets are enabling researchers to study the cancer genome in unprecedented detail. However, their sheer size and diversity presents significant challenges to downstream interpretation. While much progress has been made in automated analysis, human review enabled by intuitive and responsive visualizations remains an essential component of data analysis. The Integrative Genomics Viewer (IGV) was developed to address this need by providing researchers with an intuitive, user-friendly, interactive visualization tool for exploration of diverse genomic datasets. IGV has extensive support for both next-generation and micro-array based platforms, and for integration of these data types with clinical and phenotypic data. It provides an intuitive interface similar to online mapping tools such as Google Maps, enabling smooth zooming and panning at all resolution scales, from whole genome to base pairs. Data can be annotated, filtered, grouped, and sorted in a variety of ways. This ability to dynamically and flexibly integrate multiple different datasets, and view them at any scale, allows investigators to elucidate complex biological relationships that are not otherwise readily apparent. In this presentation we describe the IGV, with emphasis on recent features focused on data integration and interpretation, developed in close collaboration with cancer researchers. Specifically (1) a multi-locus pathway view which supports simultaneous viewing of data in multiple genomic regions defined by pathways or gene sets, (2) integration with external tools such as Mutation Assessor and PolyPhen, to provide views and data to asses the functional significance of somatic events, and (3) a flexible, interactive charting capability to enable detailed exploration of inter-dependencies between data types at a specific locus or set of loci. These features will be illustrated in the context of use cases drawn from the TCGA glioblastoma multiforme and ovarian datasets. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 3968. doi:1538-7445.AM2012-3968

The Tumor Microenvironment Strongly Impacts Master Transcriptional Regulators and Gene Expression Class of Glioblastoma

The Cancer Genome Atlas (TCGA) project has generated gene expression data that divides glioblastoma (GBM) into four transcriptional classes: proneural, neural, classical, and mesenchymal. Because transcriptional class is only partially explained by underlying genomic alterations, we hypothesize that the tumor microenvironment may also have an impact. In this study, we focused on necrosis and angiogenesis because their presence is both prognostically and biologically significant. These features were quantified in digitized histological images of TCGA GBM frozen section slides that were immediately adjacent to samples used for molecular analysis. Correlating these features with transcriptional data, we found that the mesenchymal transcriptional class was significantly enriched with GBM samples that contained a high degree of necrosis. Furthermore, among 2422 genes that correlated with the degree of necrosis in GBMs, transcription factors known to drive the mesenchymal expression class were most closely related, including C/EBP-β, C/EBP-δ, STAT3, FOSL2, bHLHE40, and RUNX1. Non-mesenchymal GBMs in the TCGA data set were found to become more transcriptionally similar to the mesenchymal class with increasing levels of necrosis. In addition, high expression levels of the master mesenchymal factors C/EBP-β, C/EBP-δ, and STAT3 were associated with a poor prognosis. Strong, specific expression of C/EBP-β and C/EBP-δ by hypoxic, perinecrotic cells in GBM likely account for their tight association with necrosis and may be related to their poor prognosis.

Abstract 4905: Patient-specific pathway analysis using PARADIGM identifies key activities in multiple cancers

Abstract Cancer is a disease of genomic perturbations that lead to dysregulation of multiple pathways within the cellular system. While common pathways are believed to be shared within specific cancer types, the mechanisms of why particular patients respond differently to treatment is not fully understood. Current -omics studies such as The Cancer Genome Atlas (TCGA) and Stand Up To Cancer (SU2C) have attempted to address this issue by using large-scale whole-genome measurements of mRNA expression, DNA copy number, and epigenetic features. Typical analysis of these measurements relies on integrating data from multiple samples to distinguish signal from noise. However, few analytical methods allow for sample-specific differences to identify features and pathways that are significant for prognosis and clinical treatment classifications. We developed a pathway inference method called PARADIGM (PAthway Recognition Algorithm using Data Integration on Genomic Models) (Bioinformatics (2010) vol. 26 (12) pp. i237) to identify patient- and sample-specific pathway activities. Previously we have shown that PARADIGM is capable of stratifying patients into clinically relevant subgroups in both TCGA ovarian serous carcinoma and glioblastoma multiforme using gene expression and copy number alteration values. We have since enhanced PARADIGM in multiple ways allowing more in-depth analysis and discovery across multiple cancer types. We have expanded the underlying pathway database previously consisting of NCI's Pathway Interaction Database to include pathways from BioCarta and Reactome, increasing the number of pathways by ten-fold and increasing the total number of features to approximately 22,000. We have refined the central dogma framework underlying all features in each pathway to support a more accurate regulatory model, allowing us to efficiently learn the strength of each interaction. Finally, PARADIGM now supports additional input data sources in the form of methylation and mutational interventions. By adding gene-level mutational information on the new regulatory model we show it is possible to distinguish between gain-of-function and loss-of-function mutations in each patient. In addition, through these enhancements we show increased ability to identify key pathway activities that effectively stratify patient cohorts. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 4905. doi:10.1158/1538-7445.AM2011-4905

Integrated genomic analyses identify ERRFI1 and TACC3 as glioblastoma-targeted genes.

The glioblastoma genome displays remarkable chromosomal aberrations, which harbor critical glioblastoma-specific genes contributing to several oncogenetic pathways. To identify glioblastoma-targeted genes, we completed a multifaceted genome-wide analysis to characterize the most significant aberrations of DNA content occurring in glioblastomas. We performed copy number analysis of 111 glioblastomas by Digital Karyotyping and Illumina BeadChip assays and validated our findings using data from the TCGA (The Cancer Genome Atlas) glioblastoma project. From this study, we identified recurrent focal copy number alterations in 1p36.23 and 4p16.3. Expression analyses of genes located in the two regions revealed genes which are dysregulated in glioblastomas. Specifically, we identify EGFR negative regulator, ERRFI1, within the minimal region of deletion in 1p36.23. In glioblastoma cells with a focal deletion of the ERRFI1 locus, restoration of ERRFI1 expression slowed cell migration. Furthermore, we demonstrate that TACC3, an Aurora-A kinase substrate, on 4p16.3, displays gain of copy number, is overexpressed in a glioma-grade-specific pattern, and correlates with Aurora kinase overexpression in glioblastomas. Our multifaceted genomic evaluation of glioblastoma establishes ERRFI1 as a potential candidate tumor suppressor gene and TACC3 as a potential oncogene, and provides insight on targets for oncogenic pathway-based therapy.

Abstract 2015: Patient-specific pathway analysis using DIGMA identifies key activities in multiple cancers

Abstract Cancer is a disease of genomic perturbations that lead to dysregulation of multiple pathways within the cellular system. While common pathways are believed to be shared within specific cancer types, the mechanisms of why particular patients respond differently to treatment is not fully understood. Current -omics studies such as The Cancer Genome Atlas (TCGA) and Stand Up To Cancer (SU2C) have attempted to address this issue by using large-scale whole-genome measurements of mRNA expression, DNA copy number, and epigenetic features. Typical analysis of these measurements relies on integrating data from multiple samples to distinguish signal from noise. However, few analytical methods allow for sample-specific differences to identify features and pathways that are significant for prognosis and clinical treatment classifications. We developed a pathway inference method called DIGMA (Directed and Integrated Graphical Modeling of Activities) to identify patient- and sample-specific pathway activities. DIGMA is capable of inferring individual gene and protein level measurements within pathways as well as overall pathway alterations specific to an individual tumor. DIGMA models each gene's “Central Dogma” within an overall pathway structure using probabilistic graphical models (PGMs). Individual -omic measurements are attached to their represented protein within the network and proteins are connected using interactions defined in curated pathway models. This representation is flexible enough to support any number of -omics measurements, which can be connected to the protein and pathway in biologically meaningful ways. Applying our method to TCGA ovarian serous carcinoma and glioblastoma multiforme samples identified pathways both significantly activated across a large majority of the cohort as well as pathways useful for classification of samples that responded well to treatment. We used random simulations to measure the significance of the pathway activities and establish a false discovery rate. These results suggest the ability to identify critical intervention points for therapeutics at a sample-specific level. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 2015.

Computational ecosystems for data-driven medical genomics

In the path towards personalized medicine, the integrative bioinformatics infrastructure is a critical enabling resource. Until large-scale reference data became available, the attributes of the computational infrastructure were postulated by many, but have mostly remained unverified. Now that large-scale initiatives such as The Cancer Genome Atlas (TCGA) are in full swing, the opportunity is at hand to find out what analytical approaches and computational architectures are really effective. A recent report did just that: first a software development environment was assembled as part of an informatics research program, and only then was the analysis of TCGA's glioblastoma multiforme multi-omic data pursued at the multi-omic scale. The results of this complex analysis are the focus of the report highlighted here. However, what is reported in the analysis is also the validating corollary for an infrastructure development effort guided by the iterative identification of sound design criteria for the architecture of the integrative computational infrastructure. The work is at least as valuable as the data analysis results themselves: computational ecosystems with their own high-level abstractions rather than rigid pipelines with prescriptive recipes appear to be the critical feature of an effective infrastructure. Only then can analytical workflows benefit from experimentation just like any other component of the biomedical research program.