Hypergraph representations of single-cell RNA sequencing data for improved cell clustering.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) analysis often uses network projections, but these unipartite approaches struggle with data sparsity and higher-order information.
• Existing methods may overestimate co-expression or underutilize sparse scRNA-seq data, limiting cell-type differentiation accuracy.

Purpose of the Study:

• To address limitations of unipartite networks in scRNA-seq analysis.
• To propose hypergraph representations for capturing higher-order relationships and handling data sparsity.
• To develop novel clustering algorithms for improved cell-type differentiation.

Main Methods:

• Representing scRNA-seq data as hypergraphs, where nodes are cells and hyperedges represent actively expressed genes.
• Introducing two hypergraph-based random walk algorithms: Dual-Importance Preference Hypergraph Walk (DIPHW) and Co-expression and Memory-Integrated Dual-Importance Preference Hypergraph Walk (CoMem-DIPHW).
• CoMem-DIPHW integrates cell-gene expression hypergraphs with gene and cell co-expression networks for enhanced clustering.

Main Results:

• CoMem-DIPHW consistently outperforms established and state-of-the-art deep learning algorithms in cell-type clustering on real-world and simulated scRNA-seq data.
• The algorithms show significant improvement on data with weak modularity.
• CoMem-DIPHW successfully annotates clusters with biologically relevant cell types.

Conclusions:

• Hypergraph representations offer a powerful framework for scRNA-seq data analysis.
• The developed CoMem-DIPHW algorithm provides a superior approach for cell-type clustering.
• This method enhances the utility of scRNA-seq data for biological discovery and cell-type annotation.