LONMF: a non-negative matrix factorization model based on graph Laplacian and optimal transmission for paired...

Area of Science:

• Single-cell multi-omics
• Computational biology
• Bioinformatics

Background:

• Single-cell sequencing technologies generate complex, noisy data.
• Integrating multi-omics data is crucial for understanding cellular heterogeneity.
• Existing methods face challenges in data integration and interpretability.

Purpose of the Study:

• To develop an effective method for integrating single-cell multi-omics data.
• To improve cell clustering and biological interpretability.
• To reveal new biological perspectives on cellular phenotypes.

Main Methods:

• Proposed LONMF, a non-negative matrix factorization algorithm.
• Combined graph Laplacian and optimal transmission for enhanced analysis.
• Applied LONMF to diverse single-cell multi-omics datasets (10X-multi-group, CITE-seq, TEA-multi-group seq).

Main Results:

• LONMF demonstrated robust performance in visualizing and clustering multi-pair single-cell multi-omics data.
• Facilitated marker characterization and gene ontology enrichment analysis.
• Achieved comparable performance to state-of-the-art methods in cell clustering.

Conclusions:

• LONMF offers superior biological interpretability compared to existing methods.
• Provides valuable biological insights for downstream single-cell multi-omics analyses.
• Enhances the understanding of cellular heterogeneity through integrated data analysis.