scapGNN: A graph neural network-based framework for active pathway and gene module inference from single-cell...

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Single-cell technologies allow multi-omics profiling but extracting functional insights is challenging.
• Existing methods struggle with the complexity and sparsity of single-cell multi-omics data.

Purpose of the Study:

• To present scapGNN, a novel graph neural network (GNN) framework for analyzing single-cell multi-omics data.
• To enable robust inference of single-cell pathway activity and identification of cell phenotype-associated gene modules.

Main Methods:

• scapGNN transforms sparse single-cell profiles into a gene-cell association network.
• Utilizes graph neural network architecture for data integration and analysis.
• Developed as a flexible and extensible R package.

Main Results:

• scapGNN demonstrates superior accuracy, robustness, and scalability compared to state-of-the-art methods.
• Successfully applied in cell denoising, batch effect removal, clustering, and trajectory inference.
• Effective in identifying cell phenotype-associated gene modules and pathway activities.

Conclusions:

• scapGNN provides a powerful new analytical platform for single-cell multi-omics research.
• Facilitates deeper understanding of cellular functions at pathway and network levels.
• Offers a flexible R package for integration into existing single-cell analysis workflows.