Colloquium: Uncovering Principles of Genome Structure – Jie Liang
Physics Colloquium Lecture
November 4, 2020
3:00 PM - 4:00 PM
Host: Jan-Hendrik Spille
Jie Liang
Center for Bioinformatics and Quantitative Biology
Dept of Bioengineering
University of Illinois at Chicago
Uncovering Principles of Genome Structure and Function from Computational Models of 3D Ensembles of Single-Cell Chromatin Conformations
3D organization of chromosomes play important roles in genome functions. Studies based on chromosome conformation capture such as Hi-C measure pairwise contacts interactions in cell populations and have uncovered broad and in-depth information on chromatin interactions at high-resolution. Single-cell Hi-C, Dip-C, and super-resolution imaging studies further revealed structural details of chromatins in individual cells, albeit with limited genomic coverage and cell numbers. Here we describe how high-resolution 3D ensemble polymer models of single-cell chromatin conformations can be constructed from Hi-C measurements, and how such models can reveal important insight into the principles of chromatin organization. Our findings include the origin of TAD-like structures, the origin of genomic looping scaling rules, the quantitative nature of chromatin heterogeneity, and how sub-populations of chromatin conformations are related to embryogenesis in Drosophila. In addition, we show that a small number of driver interactions can be predicted with minimal assumptions of physical properties and no adjustable parameters. Further, we show these interactions are sufficient to drive chromatin folding: For a diverse set of 39 highly-active transcriptional loci (480 Kb - 1.94 Mb) with ≥2 super-enhancers in human cells, the aggregation of 5x104 single-cell chromatin models obtained from 15-32 predicted driver interactions can reproduce measured population Hi-C heatmaps of 2,414-62,785 interactions (r=0.96-0.97). We further discuss how functional units of higher-order many-body interaction hubs among genomic regions can be discovered. A genome-wide investigation of many-body interactions in all highly-active transcriptional loci enabled the construction of the overall landscape of multivalent condensed hubs of super-enhancer complexes, which demonstrate significant enrichment in functional interactions (e.g., P-SE-SE and SE-SE-E). Finally, we show that principal loops of these condensed hub of interactions can be predicted through machine learning from biomarks such as increased DNase-accessibility, POLR2A-binding, and decreased H3K27me3 modifications.
Join us via Zoom: https://uic.zoom.us/j/81302083914?pwd=QjZNTisyaEk3bFVVL1dtQWdnUEdLdz09
Meeting ID: 813 0208 3914
Passcode: 00u0Rpvw
Date posted
Oct 20, 2020
Date updated
Oct 28, 2020