Finding Topologically Associating Domains

Abstract

In this thesis, we propose a new method for finding Topologically Associating Domains, which are contiguous segments of chromatin, ranging in size from thousands to millions of base pairs. These domains, which are apparent throughout most of the genome, have been postulated as being fundamental building blocks of higher-order genome structure, and being linked to the biological function of the DNA. Our method uses Hi-C interaction matrices that describe the interaction frequency between pairs of loci. The method produces a set of hierarchically nested domains, and a set of non-overlapping consensus domains — both of which can be used in further biological analyses. We made our method and domains accessible by creating three tools in the Genomic HyperBrowser. These tools can be used to create domain sets, to visualize domains with the Hi-C data, and to compare and analyse domain sets. We analyse the association between the domains and CTCF binding sites, and compare domains found in the human genome with those found in the mouse genome. We discuss how these types of analysis have been performed by others, and propose alternative ways of performing them. Our domains are similar to those found by others, but they are more self-interacting and interact less with their surroundings. Based on the strong self-interacting nature of our domains, and their association with biological features, we argue that we find a preferable set of domains.