Deep generative modeling and clustering of single cell Hi-C data.

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Understanding 3D genome conformation is crucial for gene regulation and cellular function.
• Single-cell Hi-C technologies provide insights into cell-to-cell variability in 3D chromatin organization.
• Analyzing sparse and heterogeneous single-cell Hi-C data requires advanced computational methods.

Purpose of the Study:

• To introduce scDEC-Hi-C, a novel computational framework for analyzing single-cell Hi-C data.
• To leverage deep generative neural networks for enhanced analysis of 3D genome architecture.
• To improve the clustering and imputation of single-cell Hi-C data.

Main Methods:

• Development of scDEC-Hi-C, a framework utilizing deep generative neural networks.
• Application of the framework to analyze sparse and heterogeneous single-cell Hi-C datasets.
• Comparative analysis against existing computational methods for single-cell Hi-C data.

Main Results:

• scDEC-Hi-C demonstrates superior performance in clustering single-cell Hi-C data.
• The framework shows improved accuracy in imputing missing data in single-cell Hi-C profiles.
• The generative capabilities of scDEC-Hi-C facilitate the identification of chromatin architecture differences across cell types.

Conclusions:

• scDEC-Hi-C offers a powerful new approach for single-cell Hi-C data analysis.
• The method enhances our ability to study cell-to-cell variability in 3D genome organization.
• scDEC-Hi-C is expected to advance our understanding of chromatin contact formation mechanisms.