Deep Imputation Bi-Stochastic Graph Regularized Matrix Factorization for Clustering Single-Cell RNA-Sequencing Data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) generates vast gene expression data, enabling exploration of cellular heterogeneity and function.
• Clustering scRNA-seq data is crucial for identifying cell populations, uncovering disease mechanisms, and discovering biomarkers.

Purpose of the Study:

• To present a novel method, DSINMF, for enhanced clustering of scRNA-seq data using deep matrix factorization.
• To improve the accuracy and robustness of cell population identification from complex transcriptomic datasets.

Main Methods:

• DSINMF employs a four-step approach: feature selection, dropout imputation for missing values, dimension reduction, and deep matrix factorization with bi-stochastic graph regularization.
• The method is designed to handle the inherent noise and sparsity of scRNA-seq data.

Main Results:

• DSINMF was compared against state-of-the-art algorithms using nine diverse scRNA-seq datasets.
• The proposed DSINMF method demonstrated superior performance compared to existing algorithms in clustering scRNA-seq data.

Conclusions:

• DSINMF offers an effective and robust approach for clustering single-cell RNA sequencing data.
• This method can significantly advance the understanding of cellular heterogeneity and its role in disease.