NMFLRR: Clustering scRNA-Seq Data by Integrating Nonnegative Matrix Factorization With Low Rank Representation

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell technologies offer insights into cell heterogeneity.
• Accurate cell classification is crucial for understanding biological mechanisms.
• Challenges in scRNA-seq data (high dimensionality, sparsity, noise) hinder existing clustering methods.

Purpose of the Study:

• To develop a novel computational framework for accurate cell type identification from scRNA-seq data.
• To overcome limitations of existing methods in handling noisy and complex single-cell transcriptomic profiles.

Main Methods:

• Integration of Low-Rank Representation (LRR) and Nonnegative Matrix Factorization (NMF) into a framework named NMFLRR.
• Utilizing nuclear norms for global data properties and graph regularization for local geometric information.
• Employing the Alternating Direction Method of Multipliers (ADMM) algorithm for iterative optimization and spectral clustering for final cell type prediction.

Main Results:

• NMFLRR framework successfully integrated global and local data features.
• The method achieved accurate and robust cell type classification across nine real scRNA-seq datasets.
• NMFLRR demonstrated superior performance compared to fifteen other competitive methods.

Conclusions:

• NMFLRR is a promising algorithm for accurate single-cell classification.
• The framework effectively handles the challenges posed by scRNA-seq data.
• The developed computational approach advances the field of single-cell data analysis.