WCSGNet: a graph neural network approach using weighted cell-specific networks for cell-type annotation in scRNA-seq.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) offers high resolution for cellular heterogeneity and molecular regulation analysis.
• Current cell type annotation methods often use gene expression profiles but overlook cell-specific gene interaction networks.
• This limitation can reduce the accuracy of cell type classification.

Purpose of the Study:

• To introduce WCSGNet, a novel graph neural network algorithm for automated cell type annotation.
• To leverage Weighted Cell-Specific Networks (WCSNs) for improved accuracy in cell type classification.
• To address the limitations of existing methods in capturing cell-specific gene association patterns.

Main Methods:

• Developed WCSGNet, a graph neural network algorithm utilizing Weighted Cell-Specific Networks (WCSNs).
• Constructed WCSNs based on highly variable genes to capture both gene expression and network structure.
• Employed extensive experimental validation on diverse scRNA-seq datasets.

Main Results:

• WCSGNet achieved superior cell type classification performance, ranking among top methods.
• Demonstrated robust stability across various datasets.
• Showcased a significant advantage in handling imbalanced datasets compared to existing methods.

Conclusions:

• WCSGNet effectively integrates gene expression and cell-specific network information for accurate cell type annotation.
• The algorithm provides a robust and stable solution for scRNA-seq data analysis.
• WCSGNet offers a promising approach for classifying cells, particularly in challenging imbalanced scenarios.