TADfit is a multivariate linear regression model for profiling hierarchical chromatin domains on replicate Hi-C data

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Topologically associating domains (TADs) are key organizational units within the three-dimensional genome.
• Understanding hierarchical TADs is crucial for deciphering genome architecture's role in gene regulation.
• Current methods for identifying TADs face limitations in handling complex hierarchical structures and replicate data.

Purpose of the Study:

• To develop a robust computational model for profiling hierarchical chromatin domains from Hi-C data.
• To introduce TADfit, a multivariate linear regression approach capable of identifying significant hierarchical TADs.
• To enhance the analysis of genome architecture by accommodating multiple Hi-C contact matrix replicates.

Main Methods:

• TADfit employs multivariate linear regression to model interaction frequencies in Hi-C contact matrices.
• The model considers all possible hierarchical TADs and uses Follow-The-Regularized-Leader (FTRL) for coefficient determination.
• TADfit is designed to handle multiple Hi-C replicates and identify partially overlapping TADs.

Main Results:

• TADfit demonstrates superior accuracy and reproducibility compared to existing methods.
• The model successfully identifies hierarchical TADs, including partially overlapping structures across replicates.
• The identified hierarchical TADs show significant biological relevance in genome organization.

Conclusions:

• TADfit provides an accurate and reproducible method for profiling hierarchical chromatin domains.
• The model's ability to handle replicate data offers a more comprehensive view of genome architecture.
• This approach advances the understanding of the relationship between 3D genome organization and gene regulation.