The three-dimensional organization of the genome plays a crucial role in gene regulation, and topologically associating domains (TADs) are key structural components of this organization. Reorganizations of TADs between different biological states, such as health and disease, are linked to important genomic functions. In this talk, I will present DiffDomain, a new algorithm that leverages high-dimensional random matrix theory to identify reorganized TADs using high-throughput chromosome conformation capture (Hi-C) data and scHi-C data. Benchmarking DiffDomain against existing methods demonstrates superior performance in detecting biologically relevant TAD reorganizations while maintaining lower false positive rates and higher true positive rates. Additionally, we introduce a new subtype of reorganized TADs uncovered through DiffDomain. Applying this method to both bulk and single-cell Hi-C data reveals associations betweeenTAD reorganization and structural variations and epigenomic changes, including altered CTCF binding patterns. Moreover, the method can robustly identify reorganized TADs using pseudo-bulk Hi-C data from as few as 100 cells per condition, providing insights into cell-to-cell variability and heterogeneity of TADs.

Topologically associating domain,Differential analysis,Hi-C,single-cell Hi