Biophysical Society Thematic Meeting| Les Houches 2019

Multiscale Modeling of Chromatin: Bridging Experiment with Theory

Poster Abstracts

23-POS Board 23 MANY-BODY CHROMATIN INTERACTIONS IN SUPER-ENHANCER TADS Alan Perez-Rathke 1 ; Qiu Sun 2 ; Valentina Boeva 3,4 ; Jie Liang 1 ; 1 University of Illinois at Chicago, Bioengineering, Chicago, Illinois, United States 2 Shanghai Jiao Tong University, Biomedical Engineering, Shanghai, Shanghai, China Chromatin interactions are thought to be important for gene regulation via enhancer-promoter looping as well as for critical functions such as cellular specialization. There is now emerging evidence that many-body (>2) chromatin interactions may be an important feature of super- enhancer (SE) regions - for example, condensing the SE region into a cohesive transcriptional apparatus. Chromosome conformation capture techniques such as Hi-C have greatly contributed to our understanding of the chromatin folding landscape. However, Hi-C has limitations as it only captures pairwise chromatin interactions and the interaction frequencies mostly represent population averages. Therefore, it is generally not possible to directly infer the existence of significant many-body chromatin interactions. With the goal of solving these problems, we have developed a computational model which utilizes physical properties of chromatin folding (e.g. nuclear confinement and self-avoidance) as well the experimental Hi-C data to reconstruct the corresponding ensemble of 3-D polymers. We deeply sample from a Bayesian generative model to infer the existence of significant many-body chromatin interactions in topologically associating domains (TADs) bounding SE regions. Specifically, we investigate: i) the prevalence of significant many-body chromatin interactions beyond random polymer folding; ii) the extent of enrichment of many-body interactions in SE regions; iii) which epigenetic markers are predictive of many-body interactions. Our analysis is performed on GM12878 and K562 cell lines at 5 KB resolution. We compare our many-body predictions to split-pool recognition of interactions by tag extension (SPRITE) clusters in GM12878. 3 Institut Curie, , Paris, Paris, France 4 Institut Cochin, , Paris, Paris, France

61