Joint Low Rank Representation with Symmetric Orthogonal Decomposition for Clustering of scRNA-seq Data

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) data are crucial for studying cellular heterogeneity and disease mechanisms.
• Accurate cell type identification is vital but challenging due to data sparseness, noise, and high dimensionality.
• Existing computational methods for cell type identification face limitations in prediction performance.

Purpose of the Study:

• To develop a novel computational method for accurate cell type identification from scRNA-seq data.
• To address the challenges posed by sparse, noisy, and high-dimensional scRNA-seq data.
• To improve the understanding of molecular mechanisms underlying complex diseases through precise cell subtyping.

Main Methods:

• Developed LRRS, a method integrating Low Rank Representation (LRR) and symmetric orthogonal decomposition.
• Introduced a novel orthogonal symmetric decomposition strategy for adaptive characterization of local properties.
• Utilized the Alternating Direction Method of Multipliers (ADMM) for efficient optimization of the graph model.
• Employed spectral clustering on the resulting similarity matrix for cell grouping.

Main Results:

• LRRS demonstrated effectiveness in predicting cell type composition across eleven benchmark datasets.
• The method showed superior performance compared to fourteen other state-of-the-art computational approaches.
• Evaluated performance using metrics such as prediction accuracy and normalized mutual information.

Conclusions:

• LRRS is an effective and robust computational method for cell type identification using scRNA-seq data.
• The novel decomposition strategy and optimization approach enhance prediction accuracy.
• LRRS facilitates deeper insights into cellular heterogeneity and disease-related molecular mechanisms.