Open-ComBind

Open-ComBind integrates binding interaction data from multiple ligands to improve selection of protein-ligand docking poses.

Key Features:

• Utilization of Multiple Ligand Data: Leverages binding interactions from diverse ligands associated with the same protein target to identify shared binding characteristics.
• Feature Similarity Distributions: Computes distributions of feature similarities between pairs of ligand poses by comparing near-native poses to sampled docked poses and quantifies likelihoods of observing interactions such as hydrogen bonds and hydrophobic contacts.
• Enhanced Pose Selection: Integrates feature similarity distributions with per-ligand docking scores to improve pose selection accuracy, with reported improvements of 5% for high-affinity ligands and 4.5% for congeneric series helper ligands.
• Reduction in RMSD: Reduces average ligand root-mean-square deviation (RMSD) in benchmark datasets by 9.0% relative to traditional methods.

Scientific Applications:

• Drug discovery and development: Improves the reliability of docking predictions to aid assessment of binding affinity and specificity of novel compounds and support identification of potential therapeutic candidates.

Methodology:

Analyzes feature similarity distributions across multiple ligand poses and integrates these distributions with per-ligand docking scores to refine pose selection.