Tissue-specific 3D genome organizations: structural determination, functional implication and folding principles
University of Southern California
Molecular and Computational Biology
It has become increasingly clear that the three-dimensional organization of the genome importantly influences transcriptional regulation. The development of genome-wide proximity ligation assays (such as Hi-C) has significantly expanded our understanding of spatial genome organization. However, individual 3D genome structures vary dramatically from cell to cell even within an isogenic sample. This stochastic nature may play key roles in cell differentiation and gene function. However, this structural variability poses also a great challenge to the interpretation of ensemble-average Hi-C data, especially for long-range interactions, which are relatively infrequent interactions but hold the key to shaping the global genome architecture. We present a probabilistic framework for a structure-based deconvolution of ensemble-average Hi-C maps. In particular, we generate a population of 3D genome structures that are entirely consistent with the input Hi-C data. Our approach explicitly considers the highly variable nature of genome structures, allowing the study of structural and functional states in the genome structure population. Using our population-based approach, we can not only predict structural features of genomes well, but also discover guiding principles of genome organization and propose important driving forces that shape chromosomal positioning in the nucleus. We also discuss methods of how to analyze populations of genome structures to reveal structure-function correlations. We propose a tensor-based method to extract frequent occurring spatial patterns and are able to relate these structural chromatin patterns with functional epigenomic data. Our approach facilitates structure-function mapping by delivering an atlas of spatial units along with their diverse functions in the human genome.