A divide and conquer strategy for recapitulating whole genome 3D structure using Hi-C data

Area of Science:

• Genomics
• Computational Biology
• Structural Biology

Background:

• The three-dimensional (3D) genome organization is crucial for biological functions and is studied using Hi-C assays that capture chromatin interactions.
• Current methods for 3D genome structure inference from Hi-C data fall into optimization-based and sampling-based paradigms, each with limitations.
• Optimization-based methods struggle with spatial dependencies and cell heterogeneity in bulk Hi-C data, while sampling-based methods are computationally intensive for whole-genome analysis and can miss inter-chromosomal contact information.

Purpose of the Study:

• To address the computational expense and information loss in existing 3D genome structure inference methods.
• To develop a novel computational framework that efficiently reconstructs whole-genome 3D structures from Hi-C data while accounting for data complexities.
• To improve the accuracy and efficiency of 3D genome structure modeling from genome-wide chromatin interaction data.

Main Methods:

• Proposed the truncated Random effect EXpression-cut and paste (tREX-cap) method, integrating the tREX model within a divide and conquer strategy.
• Leveraged the strengths of sampling-based approaches to account for data features like dependency, heterogeneity, over-dispersion, and sparsity.
• Implemented a strategy to enable efficient whole-genome 3D structure inference, overcoming computational bottlenecks of previous sampling-based methods.

Main Results:

• The tREX-cap method demonstrated efficient inference of whole-genome 3D structures.
• The method successfully inherited the data-feature-cognizant properties of the tREX model.
• Performance was validated through extensive simulations and analysis of real Hi-C datasets from lymphoblastoid cells and IMR90 cells.

Conclusions:

• tREX-cap provides an efficient and accurate approach for whole-genome 3D genome structure reconstruction from Hi-C data.
• The method effectively handles complexities inherent in Hi-C data, including cell heterogeneity and spatial dependencies.
• This advancement facilitates a deeper understanding of genome architecture and its functional implications.