scLINE: A multi-network integration framework based on network embedding for representation of single-cell RNA-seq...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) offers high-resolution transcriptional profiles for biomedical research.
• scRNA-seq data is characterized by high sparsity and variability due to technical limitations.
• Analyzing scRNA-seq data requires robust methods to overcome these challenges.

Purpose of the Study:

• To develop a novel method, scLINE, for learning low-dimensional representations of scRNA-seq data.
• To leverage gene-gene interaction networks and prior biological knowledge for improved data analysis.
• To enhance the accuracy and effectiveness of downstream scRNA-seq analyses.

Main Methods:

• scLINE employs a network embedding model that integrates multiple gene-gene interaction networks.
• The method incorporates prior biological knowledge to extract meaningful biological signals.
• Performance was evaluated on eight diverse single-cell datasets.

Main Results:

• scLINE demonstrated comparable or superior performance against established methods like PCA, t-SNE, and Isomap.
• The method showed strong results in internal validation metrics and clustering accuracy.
• Low-dimensional representations generated by scLINE proved effective for visualization, clustering, and cell typing.

Conclusions:

• scLINE is an effective tool for dimensionality reduction in scRNA-seq data analysis.
• The method successfully addresses data sparsity and variability by incorporating network information.
• An R package implementation facilitates the integration of scLINE into existing bioinformatics pipelines.