Graph-Regularized Non-Negative Matrix Factorization for Single-Cell Clustering in scRNA-Seq Data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) offers novel insights into cellular heterogeneity.
• Technical noise, such as dropout values, complicates scRNA-seq data interpretation.
• Accurate single-cell analysis is vital for understanding biological mechanisms.

Purpose of the Study:

• To develop an unsupervised learning framework for robust scRNA-seq data analysis.
• To address challenges in scRNA-seq data interpretation, including noise and heterogeneity.
• To enhance cell clustering and biological interpretation of scRNA-seq data.

Main Methods:

• Proposed Sc-GNNMF, an unsupervised framework utilizing graph-regularized non-negative matrix factorization (GNNMF).
• Estimated cell-cell and gene-gene sparse similarities using Laplacian kernels and p-nearest neighbor graphs (p-NNG).
• Optimized scRNA-seq data using weighted p-nearest known neighbors (p-NKN) for improved matrix decomposition.

Main Results:

• Sc-GNNMF demonstrated superior performance across 11 real scRNA-seq datasets.
• The method showed enhanced compatibility and robustness compared to existing approaches.
• Achieved excellent results in cell clustering, gene marker extraction, and pseudo-temporal analysis.

Conclusions:

• Sc-GNNMF provides an effective unsupervised learning framework for scRNA-seq data analysis.
• The approach improves the accuracy of cell type identification and biological interpretation.
• Sc-GNNMF is a robust and compatible tool for diverse scRNA-seq datasets.