MVSO-PPIS: a structured objective learning model for protein-protein interaction sites prediction via multi-view...

Area of Science:

• Computational Biology
• Bioinformatics
• Structural Biology

Background:

• Predicting protein-protein interaction (PPI) sites is crucial for understanding biological processes and reducing experimental costs.
• Current methods struggle with predicting residue subsequences at transitional boundaries, such as singular or edge structures.

Purpose of the Study:

• To develop a novel method, MVSO-PPIS, for enhanced prediction of protein-protein interaction sites, particularly focusing on transitional boundary residues.
• To improve the accuracy and structural interpretability of PPI site predictions.

Main Methods:

• MVSO-PPIS integrates a subgraph-based module and an enhanced graph attention module for feature extraction.
• An attention-based fusion mechanism combines features to capture local substructures and global dependencies.
• The model is trained to optimize prediction accuracy, edge structural consistency, and recognition of unique structural patterns.

Main Results:

• MVSO-PPIS demonstrates superior performance compared to existing baseline models on benchmark datasets.
• The method achieves higher accuracy in predicting PPI sites.
• Enhanced structural interpretability is achieved for the predicted PPI sites.

Conclusions:

• MVSO-PPIS offers a significant advancement in PPI site prediction, especially for challenging transitional boundary regions.
• The method's integrated approach provides a more comprehensive understanding of protein interactions.
• The developed method and resources are publicly available for further research.