Taba

Taba predicts binding affinity of protein-ligand complexes by integrating a mass-spring system with supervised machine-learning using crystallographic structures and binding affinity data.

Key Features:

• Binding affinity prediction: Predicts binding affinities from the atomic coordinates of protein-ligand complexes.
• Input data: Uses crystallographic structures and associated binding affinity data as training and evaluation sources.
• Representation: Models protein-ligand interactions using atomic-coordinate–based representations.
• Modeling approach: Integrates a mass-spring system with supervised machine-learning to explore interaction features.
• Scoring function exploration: Explores scoring function space to generate models tailored to individual protein systems.
• Implementation: Implemented in Python.
• Comparative performance: Can generate predictive models reported to outperform traditional scoring functions used in Molegro Virtual Docker, AutoDock4, and AutoDock Vina.

Scientific Applications:

• Computational affinity estimation: Quantitative prediction and ranking of ligand binding affinities for protein-ligand complexes.
• System-specific model generation: Creation of scoring models tailored to individual protein targets to improve prediction accuracy.
• Drug discovery and molecular interaction analysis: Support for ligand optimization and interpretation of molecular interactions in drug design workflows.

Methodology:

Integrates a mass-spring system with supervised machine-learning using crystallographic structures and binding affinity data derived from atomic coordinates; implemented in Python.