Deep representation learning of protein-protein interaction networks for enhanced pattern discovery

Area of Science:

• Bioinformatics
• Computational Biology
• Network Science

Background:

• Protein-protein interaction (PPI) networks are crucial for understanding biological system dynamics.
• Discerning complex patterns within these networks presents a significant analytical challenge.
• Existing methods often struggle with holistic node relationship characterization.

Purpose of the Study:

• To introduce a novel self-supervised network embedding framework, discriminative network embedding (DNE).
• To improve the characterization of node relationships in PPI networks, both locally and globally.
• To enhance pattern discovery and analysis in biological networks.

Main Methods:

• Developed a self-supervised network embedding framework named discriminative network embedding (DNE).
• DNE leverages contrastive learning between neighboring and distant nodes for robust representation.
• Applied DNE to analyze protein-protein interaction networks.

Main Results:

• DNE demonstrated superior performance compared to existing network embedding techniques.
• The framework effectively improved protein-protein interaction inference.
• DNE successfully identified protein functional modules within biological networks.

Conclusions:

• Discriminative network embedding (DNE) offers a robust strategy for node representation in PPI networks.
• The DNE framework provides enhanced capabilities for critical network analyses.
• This approach holds promise for advancing diverse biomedical applications.