DGCPPISP: a PPI site prediction model based on dynamic graph convolutional network and two-stage transfer learning.

Area of Science:

• Computational Biology
• Bioinformatics
• Structural Biology

Background:

• Protein-protein interactions (PPIs) are fundamental to biological processes.
• Accurate prediction of PPI sites is crucial for biology, medicine, and pharmacy.
• Enhancing deep learning model performance for PPI site prediction remains a challenge.

Purpose of the Study:

• To develop a novel model for predicting protein-protein interaction (PPI) sites.
• To leverage dynamic graph convolutional neural networks and transfer learning for improved prediction.
• To address the challenge of enhancing predictive performance in PPI site prediction.

Main Methods:

• Proposed a novel PPI site prediction model, DGCPPISP.
• Employed a dynamic graph convolutional neural network.
• Utilized a two-stage transfer learning strategy, including feature input and model training.

Main Results:

• DGCPPISP demonstrated superior performance on two benchmark datasets.
• Outperformed competing methods in F1-measure, AUPRC, and MCC metrics.
• Achieved significant performance gains over existing state-of-the-art methods like EGRET and HN-PPISP.

Conclusions:

• The DGCPPISP model is effective for PPI site prediction.
• The proposed dynamic graph convolutional network and transfer learning approach enhance predictive accuracy.
• DGCPPISP represents a significant advancement in computational prediction of protein-protein interaction sites.