NETWORK MODELLING OF TOPOLOGICAL DOMAINS USING HI-C DATA.

Area of Science:

• Genomics
• Computational Biology
• Systems Biology

Background:

• Chromosome conformation capture techniques like Hi-C reveal genome's 3D organization.
• Topologically associating domains (TADs) are key functional units within this 3D structure, regulating gene expression.
• Existing TAD detection algorithms, particularly community detection methods, often overlook the non-exchangeable nature of genomic loci.

Purpose of the Study:

• To propose a novel network model for TAD detection that accounts for the non-exchangeability of genomic positions.
• To integrate cell-type specific CTCF binding sites as covariates for enhanced TAD identification.
• To identify conserved TADs across different cell types.

Main Methods:

• Developed a network model for Hi-C data analysis that respects the sequential nature of genomic loci.
• Incorporated CTCF binding site data as biological covariates within the model.
• Utilized likelihood optimization via relaxation for efficient model fitting.
• Validated the model using simulated data and compared its performance against spectral clustering.
• Applied the model to real Hi-C data from multiple cell types.

Main Results:

• The proposed model accurately identifies TADs by considering the non-exchangeability of genomic positions.
• The model successfully integrates CTCF binding sites, improving TAD detection specificity.
• Demonstrated the model's ability to identify conserved TADs across different cell types.
• Showcased the model's advantages over traditional community detection methods using simulated data.
• Identified TADs with significant epigenetic features in real Hi-C data.

Conclusions:

• The novel network model provides a more accurate and biologically informed approach to TAD detection from Hi-C data.
• The integration of CTCF binding sites enhances the biological relevance and conservation analysis of TADs.
• This method offers a robust tool for studying genome architecture and its functional implications across cell types.