Topological identification and interpretation for single-cell epigenetic regulation elucidation in multi-tasks using...

Area of Science:

• Genomics
• Computational Biology
• Epigenetics

Background:

• Single-cell ATAC-seq (scATAC-seq) provides insights into gene regulation and cellular heterogeneity.
• Challenges include data sparsity and high dimensionality due to low sequencing depth, hindering regulatory element characterization.
• Existing methods struggle with dropout events and identifying subtle regulatory regions.

Purpose of the Study:

• To develop a novel deep graph representation learning method for single-cell chromatin accessibility data.
• To simultaneously learn data representation and perform cell clustering.
• To improve the identification and characterization of gene regulatory elements.

Main Methods:

• Developed scAGDE (single-cell Chromatin Accessibility Graph Deep autoencoder), a deep graph representation learning framework.
• Explicitly modeled the data generation process for joint representation and clustering.
• Applied scAGDE to diverse scATAC-seq datasets, including immune cells and human brain tissue.

Main Results:

• scAGDE outperformed existing methods in cell segregation, marker identification, and visualization.
• The method effectively mitigated dropout events and uncovered previously hidden accessible chromatin regions.
• Identified enhancer-like regions regulating CTLA4 and CD8A in immune cells.
• Successfully annotated cell types in human brain tissue based on cis-regulatory elements.

Conclusions:

• scAGDE is a powerful tool for analyzing single-cell chromatin accessibility data, offering superior performance over existing methods.
• The model effectively elucidates complex regulatory landscapes and identifies cell-type-specific regulatory elements.
• scAGDE advances the understanding of gene regulation and cell type diversity in complex tissues.