Robust subspace structure discovery for cell type identification in scRNA-seq data.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) provides high-resolution cellular heterogeneity insights.
• Accurate cell type identification is crucial but challenging due to data noise, sparsity, and high dimensionality.
• Existing clustering methods struggle with the complexities of scRNA-seq data.

Purpose of the Study:

• To introduce a novel deep subspace clustering method for robust cell type identification in scRNA-seq data.
• To enhance the accuracy and interpretability of cell clustering by addressing data limitations.
• To improve the characterization of cellular heterogeneity using advanced computational techniques.

Main Methods:

• A novel deep subspace clustering approach utilizing a self-representation learning framework.
• Integration of a structure-guided strategy with an optimal transport algorithm for optimization.
• Application to 18 real-world scRNA-seq datasets for validation.

Main Results:

• The proposed method effectively captures reliable subspace structures from noisy and sparse scRNA-seq data.
• Demonstrated superior performance compared to state-of-the-art methods across multiple datasets.
• Achieved higher accuracy and improved interpretability in cell type identification.

Conclusions:

• The novel deep subspace clustering method offers a robust solution for scRNA-seq data analysis.
• This approach significantly enhances cell type identification accuracy and biological interpretation.
• It represents a valuable advancement in computational methods for single-cell genomics.