GHGPR-PPIS: A graph convolutional network for identifying protein-protein interaction site using heat kernel with...

Area of Science:

• Computational Biology
• Bioinformatics
• Structural Biology

Background:

• Accurate identification of protein-protein interaction sites (PPIS) is crucial for understanding protein function and disease mechanisms.
• Existing computational PPIS prediction methods show limitations in predictive accuracy.
• Traditional experimental methods for PPIS identification are labor-intensive and time-consuming.

Purpose of the Study:

• To develop a novel, highly accurate computational model for predicting protein-protein interaction sites (PPIS).
• To enhance the performance of graph-based models for PPIS prediction through advanced techniques.
• To provide a more reliable and efficient tool for PPIS identification in molecular biology.

Main Methods:

• Proposed GHGPR-PPIS, a graph-based computational model integrating GraphHeat and Generalized PageRank (GHGPR).
• Incorporated an edge self-attention feature processing block within the GHGPR framework to improve performance.
• Utilized t-SNE dimensionality reduction and clustering for interpretability analysis.

Main Results:

• GHGPR-PPIS outperformed all competing state-of-the-art models on a benchmark test set.
• Demonstrated superior generalization performance and practical applicability compared to AGAT-PPIS on independent test sets.
• Interpretability analysis confirmed the effectiveness of GHGPR-PPIS across different model stages.

Conclusions:

• GHGPR-PPIS represents a significant advancement in computational prediction of protein-protein interaction sites.
• The model offers improved accuracy and generalizability, addressing limitations of current methods.
• The findings support the practical applicability of GHGPR-PPIS in biological research and disease mechanism studies.