Bayesian Estimation of Three-Dimensional Chromosomal Structure from Single-Cell Hi-C Data.

Area of Science:

• Genomics and Molecular Biology
• Computational Biology and Bioinformatics
• Structural Biology

Background:

• Inferring three-dimensional (3D) chromosome structure is crucial for understanding genome regulation.
• Bulk Hi-C data provides population-averaged structural information but misses cell-to-cell variability.
• Single-cell Hi-C data offers insights into individual cell structures but suffers from sparsity.

Purpose of the Study:

• To develop a computational method for robust 3D chromosome structure inference from sparse single-cell Hi-C data.
• To integrate bulk Hi-C data as prior information to enhance single-cell structural inference.
• To investigate the impact of prior strength on the landscape of inferred 3D structures.

Main Methods:

• Developed a Bayesian multiscale approach, termed Structural Inference via Multiscale Bayesian Approach (SIMBA).
• Incorporated bulk Hi-C data and regularization terms as priors within the Bayesian framework.
• Analyzed the solution space by varying prior strength and assessed solution clustering for single-cell data.

Main Results:

• The Bayesian multiscale approach successfully infers 3D chromosome structures from sparse single-cell Hi-C data.
• Integration of bulk Hi-C data as prior information significantly improves structural inference accuracy.
• Demonstrated that solutions derived from the same cell cluster together, validating the method's consistency.

Conclusions:

• The developed Bayesian multiscale method provides a robust framework for 3D chromosome structure reconstruction from single-cell Hi-C.
• This approach effectively leverages complementary information from bulk and single-cell Hi-C data.
• The method facilitates the study of inter-cellular variability in chromosome architecture.