Abstract
DNA is heterogeneously packaged into chromatin, which is further organized into topologically associating domains (TADs) with sharp boundaries. These boundary locations are critical for genome regulation. Here, we explore how the distribution of DNA-packing density across chromatin affects the TAD boundary locations. We develop a polymer-physics-based model that utilizes DNA accessibility data to parameterize DNA-packing density along chromosomes, treating them as heteropolymers, and simulates the stochastic folding of these heteropolymers within a nucleus to yield a conformation ensemble. Such an ensemble reproduces a subset (over 60%) of TAD boundaries across the human genome, as confirmed by Hi-C data. Additionally, it reproduces the spatial distance matrices of 2-Mb genomic regions provided by FISH experiments. Furthermore, our model suggests that utilizing DNA accessibility data alone as input is sufficient to predict the emergence and disappearance of key TADs during early T cell differentiation. We show that stochastic folding of heteropolymers in a confined space can replicate both the prevalence of chromatin domain structures and the cell-to-cell variation in domain boundary positions observed in single-cell experiments. Furthermore, regions of lower DNA-packing density preferentially form domain boundaries, and this preference drives the emergence of TAD boundaries observed in ensemble-averaged Hi-C maps. The enrichment of TAD boundaries at CTCF binding sites can be attributed to the influence of CTCF binding on local DNA-packing density in our model. Collectively, our findings establish a strong link between TAD boundaries and regions of lower DNA-packing density, providing insights into the mechanisms underlying TAD formation.