Identifying TAD-like domains on single-cell Hi-C data by graph embedding and changepoint detection

Area of Science:

• Genomics
• Computational Biology
• Epigenetics

Background:

• Topologically associating domains (TADs) are key structural units of the 3D genome.
• Single-cell TAD-like domains are crucial for understanding bulk TAD formation and function.
• Identifying TADs in sparse single-cell Hi-C data is computationally challenging.

Purpose of the Study:

• To introduce scKTLD, an computational tool for identifying TAD-like domains in single-cell Hi-C data.
• To analyze the organization and conservation of TAD-like domains at the single-cell level.
• To investigate the regulatory significance of identified TAD boundaries.

Main Methods:

• scKTLD models Hi-C contact matrices as graphs.
• It employs sparse matrix factorization and spectral propagation for dimensionality reduction.
• Kernel-based changepoint detection is used to identify TAD-like domains in the embedded space.

Main Results:

• scKTLD outperforms existing methods on sparse single-cell Hi-C data.
• The study demonstrates conserved TAD-like domain boundaries across single cells.
• High-frequency boundaries are enriched for architectural proteins and epigenetic marks, correlating with bulk TADs.

Conclusions:

• scKTLD provides a robust method for TAD identification in sparse single-cell Hi-C data.
• TAD boundaries exhibit both conservation and heterogeneity at the single-cell level.
• Conserved TAD boundaries play significant roles in genome regulation.