DiffDomain enables identification of structurally reorganized topologically associating domains

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Topologically associating domains (TADs) are fundamental to mammalian genome's 3D organization.
• Dynamic TAD reorganization is linked to genome function in health and disease.
• Current computational methods for identifying reorganized TADs are limited.

Purpose of the Study:

• To introduce DiffDomain, a novel algorithm for identifying structurally reorganized TADs.
• To leverage high-dimensional random matrix theory for analyzing chromosome conformation capture (Hi-C) contact maps.
• To provide a robust computational tool for comparative analysis of TADs.

Main Methods:

• Developed DiffDomain algorithm based on high-dimensional random matrix theory.
• Utilized high-throughput chromosome conformation capture (Hi-C) contact maps as input.
• Compared DiffDomain performance against existing methods using multiple Hi-C datasets.

Main Results:

• DiffDomain demonstrated superior performance in false and true positive rates compared to alternative methods.
• The algorithm successfully identified a novel subtype of reorganized TADs.
• Analysis revealed associations between reorganized TADs and structural variations/epigenomic changes (e.g., CTCF binding).
• DiffDomain effectively identified reorganized TADs in single-cell Hi-C data from mouse neuronal development with high reproducibility.
• The method uncovered differential variability in TADs at both cell-population and cell-to-cell levels.

Conclusions:

• DiffDomain is a statistically robust and accurate method for identifying reorganized TADs.
• The algorithm enhances the comparative analysis of TADs using both bulk and single-cell Hi-C data.
• DiffDomain offers significant advancements in understanding genome organization dynamics across different biological states.