DGCddG: Deep Graph Convolution for Predicting Protein-Protein Binding Affinity Changes Upon Mutations.

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Accurate prediction of protein--protein interactions is vital for understanding biological mechanisms and advancing drug design.
• Amino acid mutations can significantly alter protein binding affinity, impacting function and therapeutic potential.

Purpose of the Study:

• To develop a novel deep graph convolution (DGC) network-based framework, DGCddG, for predicting changes in protein-protein binding affinity following mutations.
• To provide a computational tool that enhances understanding of mutation effects on protein interactions.

Main Methods:

• Utilized a deep graph convolution network (DGC) to extract contextualized residue representations from protein complex structures.
• Employed a multi-layer perceptron to correlate DGC-mined mutation site features with binding affinity changes.
• Validated the DGCddG framework on multiple datasets, including single and multi-point mutations.

Main Results:

• DGCddG demonstrated robust performance in predicting binding affinity changes for both single and multi-point mutations.
• The model achieved superior results in blind tests predicting changes in angiotensin-converting enzyme 2 (ACE2) binding with SARS-CoV-2.
• These findings suggest DGCddG's potential utility in identifying mutations that affect viral-host interactions.

Conclusions:

• The DGCddG framework offers an effective approach for predicting mutation-induced changes in protein binding affinity.
• This method can significantly contribute to protein function studies and accelerate drug discovery and antibody design.
• The framework's performance on SARS-CoV-2 related data highlights its relevance in infectious disease research.