Two-Stage Multi-View Graph Spectral Clustering for Single-Cell RNA-Seq Data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) enables gene expression analysis at the cellular level but faces challenges like high dimensionality and data sparsity.
• Existing clustering methods struggle to fully leverage the rich information within scRNA-seq data, particularly the structural relationships between cells.

Purpose of the Study:

• To develop an advanced multi-view learning framework to comprehensively characterize scRNA-seq data from diverse perspectives.
• To address the limitations of single characteristic spaces in scRNA-seq data analysis and improve cell type distribution studies.

Main Methods:

• Constructed multiple characteristic spaces for scRNA-seq data.
• Employed a two-stage multi-view learning approach involving weighted and structural learning of similarity graphs.
• Utilized the alternating direction multiplier method for optimizing attribute and structure graphs.

Main Results:

• The proposed Multi-View Graph based Structure Clustering (MVGSC) method was validated on eight scRNA-seq datasets.
• MVGSC demonstrated significantly superior performance compared to both single-view and other multi-view clustering methods.
• The two-stage approach effectively captured complex relationships and differences across multiple data views.

Conclusions:

• Two-stage multi-view learning enhances model accuracy by capturing complex data relationships.
• This approach improves generalization by effectively utilizing consistency and complementary information across multiple views in scRNA-seq data.