Join Persistent Homology (JPH)-Based Machine Learning for Metalloprotein-Ligand Binding Affinity Prediction.

Area of Science:

• Computational chemistry
• Bioinformatics
• Structural biology

Background:

• Metalloproteins are vital in biological processes like respiration and drug metabolism.
• Targeting metalloproteins is crucial for drug design and discovery.
• Topological data analysis (TDA) shows promise in drug discovery applications.

Purpose of the Study:

• To introduce join persistent homology (JPH) for the first time.
• To develop JPH-based machine learning models for metalloprotein-ligand binding affinity prediction.
• To offer a more comprehensive molecular characterization using JPH descriptors.

Main Methods:

• Developed a novel join persistent homology (JPH) method using multistage filtration.
• Generated JPH-based molecular descriptors for enhanced topological information.
• Integrated JPH descriptors with a gradient boosting tree (GBT) model for affinity prediction.

Main Results:

• The JPH-GBT model achieved superior performance in metalloprotein-ligand binding affinity prediction.
• JPH descriptors provided a more comprehensive characterization of molecular structures.
• The model outperformed all existing methods on the PDBbind-v2020 benchmark dataset.

Conclusions:

• Join persistent homology (JPH) is a powerful tool for characterizing molecular structures and functions.
• JPH-based machine learning models hold significant potential for drug discovery.
• This approach advances the prediction of metalloprotein-ligand interactions.