Three-dimensional regulation: Establishing novel linkages between non-coding genetic variation and target genes

Abstract

In the human genome, 98% of the DNA is in non-gene coding regions. While these regions do not express genes, a mounting number of studies have shown that they are crucial to the maintenance of chromosome structure and in the regulation of gene expression. Although large epigenomics projects were established to functionally annotate non-coding regions, the comprehensive linkages between these regions and their target genes remain unknown. The human genome folds into hierarchical three-dimensional (3D) structure, bringing distantly regulatory elements into close proximity, leading to the formation of 3D chromatin physical interactions and playing an important role in the complex gene regulation network. Using chromatin interaction information, we can connect functional non-coding regions to their target genes to reveal novel regulation mechanisms. In Chapter 1, we reviewed current existing approaches to prioritise functional interactions from Hi-C data, the state-of-the-art data type used to study chromatin interactions, and categorised them into three classifications, including structuralbased methods, statistical model-based methods and data integration methods. Chapter 2 described the computational procedures of analysing Hi-C datasets, and introduced: HiC-QC, a tool that extracting summary statistics to perform quality control with Hi-C libraires; HiC-interactionmap and integration-tracks plot, tools to offer visualisation for Hi-C data integration. Additionally, aligners BWA and Bowtie2, were compared for their performance of mapping Hi-C data. Using type 1 diabetes (T1D) and regulatory T cells (Treg) as a disease-cell type model, based on data integration of Treg-specific Hi-C interactions and other epigenomics information, Chapter 3 established a filtering workflow called 3DFAACT-SNPs to link genetic variants that are associated with T1D to the loss of immune tolerance in Treg. Using this workflow, we identified 36 SNPs with plausible Treg-specific mechanisms of action contributing to T1D, linking 119 novel interacting regions. We demonstrated that it is possible to prioritise SNPs that contribute to disease based on regulatory function and illustrate the power of using chromatin interactions to connect non-coding SNPs to disease mechanisms. Lastly, Chapters 4 and 5 launch the statistically significant interaction profiling of 51 human cell lines and primary tissues from 173 public Hi-C datasets using a statistical model from MaxHiC, followed by investigating the uniqueness, distancing preference and the associated genes of the cell/tissue-specific interactions. We also identified interaction “hot zones”, regions with chromatin interactions observed across many cells and tissues. Using global and local enrichment analysis and a comparison to frequent interacting regions, we demonstrated the structural and regulatory functionality of the hot zones. We further comprehensively annotated chromatin interactions into 66 interaction classes, cataloguing potentially regulatory functional interactions for different cells and tissues. Finally, we revealed cell/tissue-specific 3D regulatory regions that are enriched with super-enhancers and overlapped with expression quantitative trait loci (eQTLs). Overall, using data integration and statistical models to prioritise functional chromatin interactions, this work produced novel computational tools and pipelines and generated valuable resource for the investigation of genome structure, demonstrating the power of using chromatin interactions to discover novel mechanisms in the genome and revealing novel linkages between noncoding DNA to traits/diseases.