Clustering scRNA-seq data with the cross-view collaborative information fusion strategy.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) provides high-throughput, cellular-resolution gene expression data.
• Cell clustering is crucial for scRNA-seq data analysis but challenged by data sparsity and high dimensionality.
• Integrating gene expression with cell structure data for clustering is an underexplored area.

Purpose of the Study:

• To develop a novel algorithm, scCFIB, for robust cell clustering in scRNA-seq data.
• To leverage the information bottleneck (IB) principle for efficient processing of high-dimensional, sparse scRNA-seq data.
• To integrate multi-view features, including gene expression and potentially cell structure, for improved clustering.

Main Methods:

• scCFIB employs an information bottleneck (IB)-based approach for clustering.
• A cross-view fusion strategy is utilized, creating two distinct feature spaces from original data.
• The algorithm formulates cell clustering as a target loss function within the IB framework with sequential optimization.

Main Results:

• scCFIB demonstrates superior performance in cell clustering tasks on diverse scRNA-seq datasets.
• The algorithm effectively handles the challenges of high-dimensional and sparse scRNA-seq data.
• Benchmarking confirms scCFIB's robustness compared to existing methods.

Conclusions:

• scCFIB offers a powerful and efficient solution for cell clustering in scRNA-seq data analysis.
• The IB-based approach and cross-view fusion strategy enhance clustering accuracy and robustness.
• The developed algorithm advances the analysis of complex single-cell genomics data.