Reconstruction of 3D Genome Architecture from Chromatin Conformation Capture Data

University Of California, San Francisco, San Francisco CA

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): It is widely recognized that the three dimensional (3D) architecture of eukaryotic chromatin plays critical roles in nuclear and cellular function. Indeed, such relationships constitute one of the key threads of the just published (09/06/12) suite of Nature papers from the ENCODE consortium. However, until a few years ago, observing / inferring, 3D structure at even modest resolutions was problematic, in part because genomes are highly condensed. Recently devised molecular techniques are changing this situation. In particular, the development of novel genome-scale assays, such as chromatin conformation capture (CCC), has enabled elicitation of chromatin contacts. These techniques have already provided insight into chromatin organization at unprecedented resolutions, and permitted exploration of the downstream influence of such organization on a variety of biological. Notably, gene regulation and cancer-driving gene fusions are believed to be strongly influenced by 3D organization. Accordingly, obtaining high resolution 3D reconstructions of genome architecture is a compelling biological quest. However, most analysis of CCC data has focused on the one dimensional (1D) contact level, with appreciably less effort directed toward evaluating accuracy and reproducibility of 3D reconstructions, and deploying such structures to analyze consequent biological processes. The overarching hypothesis to be addressed in this proposal is that chromatin contact data can be reliably used to determine 3D structures of genomes and to assess downstream relationships with biological function. To test this hypothesis we will develop new, and refine existing, reconstruction algorithms, and will undertake a systematic evaluation of their performance and operating characteristics. The only genome-scale algorithms developed employ constrained optimization which, on account of high-dimensionality, is computationally burdensome and can be trapped in local optima. Accordingly, we will investigate alternate algorithms. Our preliminary work indicates that reproducibility under plausible perturbations to data and constraint inputs is qualitatively poor. We will devise metrics to quantify 3D agreement. While reproducibility under perturbations can be assessed using computational and statistical tools, appraising accuracy requires using targeted confirmatory assays, as will be deployed by our wet-lab collaborators. Further, we will apply our compendium of 3D reconstructions to explore a series of biologic questions relating to proximities of functional elements.