A robust semi-supervised NMF model for single cell RNA-seq data.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables cellular heterogeneity analysis.
• Clustering is crucial for scRNA-seq data analysis, aiding in cell function understanding.
• Existing unsupervised Nonnegative Matrix Factorization (NMF) methods overlook valuable gene function prior knowledge.

Purpose of the Study:

• To develop a robust and semi-supervised NMF (rssNMF) model for scRNA-seq data clustering.
• To integrate known cell marker genes as prior information into the clustering process.
• To enhance the accuracy and biological relevance of cell clustering in scRNA-seq data.

Main Methods:

• Proposed a robust and semi-supervised NMF (rssNMF) model.
• Introduced a new variable to absorb data noise.
• Incorporated marker genes as prior information via a graph regularization term.

Main Results:

• The rssNMF model demonstrated superior performance compared to NMF, KMeans, and Hierarchical Clustering on twelve scRNA-seq datasets.
• Biological significance analysis confirmed rssNMF's ability to identify key cell subclasses and latent biological processes.
• This study presents the first method to integrate prior knowledge into scRNA-seq data clustering.

Conclusions:

• The developed rssNMF model offers a significant advancement in scRNA-seq data analysis.
• Incorporating prior knowledge, specifically marker genes, substantially improves clustering accuracy.
• This approach provides a powerful tool for dissecting cellular heterogeneity and biological functions.