Spatom: a graph neural network for structure-based protein-protein interaction site prediction

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Identifying protein-protein interaction (PPI) sites is crucial for understanding biological processes.
• Current computational methods face challenges in accurately predicting these interaction sites.

Purpose of the Study:

• To introduce Spatom, a novel computational framework for precise protein-protein interaction site prediction.
• To improve the accuracy and reliability of identifying protein binding interfaces.

Main Methods:

• Representing protein structures as weighted digraphs to capture spatial residue contacts.
• Employing weighted digraph convolution for aggregating local and global spatial information.
• Utilizing an improved graph attention layer to ensure prediction continuity.

Main Results:

• Spatom achieved superior performance compared to existing methods on diverse protein-protein complexes.
• The framework accurately identified interaction interfaces and key binding hotspots in case studies.
• Demonstrated significant improvements in predicting continuous interaction regions.

Conclusions:

• Spatom offers a robust and accurate solution for protein-protein interaction site prediction.
• The framework has the potential to advance the understanding of protein interactions.
• Spatom is expected to aid in the design of targeted drugs for protein binding.