BINDER achieves accurate identification of hierarchical TADs by comprehensively characterizing consensus TAD boundaries

Abstract

As a crucial chromatin structure, hierarchical TADs play important roles in epigenetic organization, transcriptional activity, gene regulation, and cell differentiation. Currently, it remains a highly challenging task to accurately identify hierarchical TADs in a computational manner. The key bottleneck for existing TAD callers lies in the difficulty on prediction of precise TAD boundaries. We solve this problem by introducing a novel algorithm, called BINDER, which conducts a boundary consensus approach, and then precisely locate hierarchical TAD boundaries by developing a multifaceted boundary characterization strategy. After comparison with other leading TAD callers, BINDER shows great improvement in identifying hierarchical TADs and exhibits the strongest robustness with ultrasparse data, which fully indicates the importance of boundary identification in calling hierarchical TADs. Applying BINDER to experimental data and mouse hematopoietic cases, we find that the hierarchical TADs identified by BINDER show strong biological relevance with epigenetic organization, transcriptional activity, DNA motifs, and coregulation during cellular differentiation. BINDER discovers differences in the enrichment of two specific transcription factors, CHD1 and CHD2 at TAD boundaries with different hierarchies. It also observes variations in gene expression of TADs with different hierarchies during cellular differentiation.