AutoDockFR: Advances in Protein-Ligand Docking with Explicitly Specified Binding Site Flexibility

Area of Science:

• Computational chemistry and structural biology
• Drug discovery and development

Background:

• Modeling receptor flexibility is crucial for accurate structure-based drug design but remains a significant challenge.
• Existing automated docking tools often struggle with receptor flexibility, leading to inaccurate ligand binding predictions.
• The computational complexity and increased false positives associated with flexible receptor modeling require specialized algorithms and scoring functions.

Purpose of the Study:

• To develop and validate a novel computational docking engine, AutoDockFR (AutoDock for Flexible Receptors), capable of modeling receptor flexibility.
• To address the challenges of increased search space and false positives in flexible receptor docking.
• To improve the efficiency and reliability of ligand docking into flexible receptor targets.

Main Methods:

• Developed AutoDockFR, a new docking engine incorporating a novel Genetic Algorithm (GA) and a customized scoring function based on AutoDock4.
• Explicitly modeled receptor flexibility by defining a set of receptor side-chains a priori.
• Validated AutoDockFR using the Astex Diverse Set, SEQ17 (apo-holo receptor pairs), and CDK2 (apo receptor with multiple inhibitors) datasets.

Main Results:

• AutoDockFR demonstrated increased efficiency and reliability of its GA compared to AutoDock4.
• Significantly higher success rates in cross-docking ligands into apo receptors requiring side-chain conformational changes: 70.6% (SEQ17) and 76.9% (CDK2) for AutoDockFR vs. 35.3% and 61.5% for AutoDock Vina, respectively.
• AutoDockFR outperformed AutoDock Vina in the number of top-ranking solutions and accurately recreated atomic interactions in docked complexes.

Conclusions:

• AutoDockFR effectively models receptor flexibility, leading to improved accuracy and success rates in automated molecular docking.
• The developed GA and scoring function successfully address the challenges of large search spaces and false positives in flexible docking.
• Down-weighting receptor internal energy further enhances pose ranking, and AutoDockFR exhibits linear runtime scaling with added flexibility.