A change-point approach to identify hierarchical organization of topologically associated domains in chromatin interaction

The identification of spatial and temporal three-dimensional (3D) genome architecture is critically important in understanding genomic functionality, gene regulation and transcription. In recent years, a higher throughput, higher resolution method of chromosome conformation capture (Hi-C) has been used to investigate genome-wide mapping of chromatin interactions, and reveals that genome is partitioned into topologically associating domains (TADs) with hierarchical structures. To identify hierarchical organization of TADs in Hi-C data, we propose generalized likelihood ratio tests for change-points in Hi-C matrices and study their asymptotic distributions. Based on these tests, we further develop a top-down and bottom-up procedure to decipher TAD structures from Hi-C contact profile. Comprehensive tests of our method on both simulated data and real Hi-C data of multiple human cell lines demonstrated that our approach can well detect hierarchical TADs organizations and show superior ability in estimating hierarchical TAD than existing methods. Our estimations on multiple cell lines show that the hierarchical TAD structures in active chromosomal regions are significantly more than those in repressive regions, indicating more precise controlling and regulation within active regions of human chromosomes.