Insights into the Regulatory Activities of Chromatin ThroughComputational Analyses of Whole-Genome Data

In the last decade, advances in DNA sequencing technology have enabled the routine generation of whole-genome datasets that profile chromatin composition and conformation. Quantitative analyses of these data have required the development of innovative computational data analysis methodologies. These analysis techniques have been used to generate novel biological hypotheses about the regulatory activities of chromatin, which can be experimentally validated. This dissertation presents several case studies of such analyses. In the first study, machine learning models, which predict transcription as a function of multivariate histone modification levels, are used to predict the association of symmetrically dimethylated arginine 3 on histone H4 (H4R3me2s) with transcriptional repression. Methodological approaches to constructing similar models are also explored in depth. In the next study, analyses of changes in a panel of histone modifications during the epithelial-mesenchymal transition (EMT) reveal a high degree of regulatory coordination among genes within distinct functional classes and pathways. Changes at enhancers associated with these genes also show coordination with respect to transcription factor binding. These observations lead to the hypothesis that histone modifications enable and sustain transcriptional feedback loops distinctly associated with each phenotypic endpoint in EMT. In the final study, a novel approach for analyzing unbiased chromatin interaction data (Hi-C data) is presented. This approach utilizes network analysis techniques to infer conformational features of the genome. Using these techniques, assessments are made of the degree of hierarchical organization in the budding yeast genome. Furthermore, a novel correlation between replication timing and degree of inter-chromosomal interactions is observed. The studies presented in this dissertation demonstrate the utility of computational data analysis in generating novel systems-level hypotheses about the regulatory behavior of chromatin. Since many of the analysis techniques presented in this work have not been otherwise applied to chromatin data, many domain-specific technical considerations are also discussed. This dissertation provides a variety of novel insights into the regulatory activities of chromatin; and perhaps more importantly, it provides several analytical frameworks for distilling systems-level insights from whole-genome chromatin data.