Spectral clustering based on learning similarity matrix.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) provides high-resolution gene expression data.
• Clustering cells based on gene expression is crucial for identifying cell types in scRNA-seq analysis.

Purpose of the Study:

• To develop a novel spectral clustering framework for improved cell type identification in scRNA-seq data.
• To leverage multiple similarity matrices and impose sparse structures for robust clustering.

Main Methods:

• Introduced a spectral clustering framework incorporating sparse structures on a target matrix.
• Utilized multiple doubly stochastic similarity matrices to learn an integrated similarity matrix.
• Employed the Alternating Direction Method of Multipliers (ADMM) algorithm to solve the non-convex optimization problem.

Main Results:

• The proposed method accurately and robustly identifies cell clusters in both simulated and real scRNA-seq datasets.
• Demonstrated the convergence of the ADMM algorithm for solving the non-convex problem.
• The framework effectively imposes desired sparse structures on the target matrix.

Conclusions:

• The novel spectral clustering framework offers a robust approach for cell type identification in scRNA-seq data.
• The method's performance was validated on diverse datasets, highlighting its accuracy.
• The source code is publicly available for broader application.