Hom-Complex-Based Machine Learning (HCML) for the Prediction of Protein-Protein Binding Affinity Changes upon...

Area of Science:

• Computational Biology
• Biophysics
• Artificial Intelligence

Background:

• Protein-protein interactions (PPIs) are fundamental to cellular processes, disease mechanisms, and therapeutic development.
• Artificial intelligence (AI) models show promise for PPI analysis, but efficient molecular representation remains a challenge.

Purpose of the Study:

• To develop a novel Hom-complex-based representation and machine learning models for predicting PPI binding affinity changes upon mutation.
• To introduce a new method for multiscale characterization of PPIs using persistent homology and persistent Euler characteristic.

Main Methods:

• Generation of Hom-complexes from graph representations of protein-protein complexes.
• Utilizing persistent homology and persistent Euler characteristic as molecular descriptors.
• Employing gradient boosting tree (GBT) machine learning models for prediction.

Main Results:

• The proposed Hom-complex-based model demonstrates superior performance in predicting PPI binding affinity changes compared to existing methods.
• Systematic testing on SKEMPI and AB-Bind datasets validates the model's effectiveness.
• The approach offers a powerful new tool for PPI analysis.

Conclusions:

• The Hom-complex-based representation provides an efficient featurization strategy for AI-based PPI models.
• This novel method holds significant potential for advancing the understanding and manipulation of PPIs.
• The model can aid in the analysis and design of targeted therapeutics, including antibodies for viruses like SARS-CoV-2.