Computational analyses of DNA methylation and gene expression for the molecular profiling of disease states

Assessing patterns of molecular entities such as DNA or RNA can contribute to an understanding of disease and describe signatures associated with clinical outcomes. The development of high throughput platforms such as microarrays and sequencing technologies allows the comprehensive molecular characterization of disease samples. These molecular profiling approaches generate large volumes of data that require the implementation of scalable computational approaches for analysis and data management. This dissertation focuses on computational analysis methods for DNA methylation and gene expression data and their application in molecular profiling studies to characterize disease states.;First, the development of computational analysis pipelines for gene expression and DNA methylation sequencing datasets is presented. These analysis pipelines are implemented in GobyWeb, a user-friendly integrated analysis suite, developed in the Campagne laboratory, that supports the analysis and management of high throughput sequencing data. Second, we profiled DNA methylation in a mouse model of anxiety to investigate the hypothesis that an adverse maternal environment characterized by a maternal serotonin receptor knockout is associated with the anxiety phenotype in adult mice raised in this environment. Primarily we found that genes encoding cell adhesion molecules and neurotransmitter receptor genes were aber-raptly methylated in mice raised by mothers with a serotonin receptor deficit (either full knock-outs, or heterozygotes). Many of the aberrantly methylated genes have been previously implicated in anxiety.;Finally we present the application of DNA methylation profiling to identify molecular subtypes of Diffuse Large B Cell Lymphoma (DLBCL). We carried out unsupervised clustering of DLBCLs based on how variable the genome-wide methylation profile compared to normal germinal center B cells and identified six DNA methylation-based clusters. The novel clusters are characterized by aberrant methylation of genes involved specific biological pathways such as cytokine-mediated signaling, ephrin signaling and pathways associated with apoptosis and cell cycle regulation. We found that the magnitude of methylation changes is significantly associated with survival outcomes in this cohort. This dissertation concludes with a discussion on future directions and perspectives of this work.