Persistent homology-based multiscale Hi-C analysis uncovers TADs-mediated gene regulation.

Topologically associating domains (TADs), computationally defined 3D chromatin structures, play pivotal roles in gene expression regulation and genome organization. However, their structural basis and functional implications remain contentious. The interplay between TADs and other multiscale 3D chromatin structures has not been comprehensively computationally characterized. The extent and regulatory scope of TAD-mediated gene expression continue to be debated. Multiscale analyses spanning hierarchical 3D chromatin structures hold promise for establishing unified structural-functional frameworks for TADs. Here, we developed topoHiC-a computational framework enabling simultaneous identification of TADs and chromatin loops from Hi-C contact matrices via persistent homology. Benchmarking analyses demonstrated improved performance of topoHiC compared to four previously published loop callers and six TAD callers. We applied topoHiC to explore multiscale 3D chromatin remodeling and showed that spatial proximity relative to TAD boundaries could modulate differential gene expression probability. We proved the potential capacity of multiscale 3D structures to predict differential gene expression through a generalized regression model. Our findings demonstrated the functional roles of TADs as architectural compartmentalization through establishing structural relationships between TADs and chromatin loops and exploring inside genes and truncated genes respectively.