Molecular docking is widely used in ligand discovery, even though it makes large errors. Here we develop model cavity sites to investigate specific problems in docking, and use the results to guide the improvement of the methods. To demonstrate relevance, and to ask questions that cannot be addressed in the small model cavities, we extend these studies to biologically relevant targets. These, in turn, inform our development of new methods.
Molecular docking is now widely used for ligand discovery. The technique makes many approximations and though it has had noteworthy successes, it is neither fully reliable nor do we understand why it fails when it does. Here we develop new docking methods that are tested in six model protein cavities engineered for docking. Each is small (150 to 200 Å3), buried from solvent, and dominated by a single term: from a fully hydrophobic cavity, to the same cavity with a single hydrogen-bond acceptor, to a cavity dominated by a single aspartate, and intermediates between. In these cavities particular interactions may be isolated and failed predictions are as illuminating as confirmed ones. Against each cavity we screen a library of 500,000 small molecules, testing predicted ligands for affinity (e.g., by ITC) and geometric fidelity, by x-ray. These studies are extended to several tru biological targets to ensure relevance. The specific aims are: 1. Development of new docking algorithms. We match methods development with experimental testing, focusing on areas where pragmatic improvement may be anticipated. A. Sampling receptor flexibility while weighting introduced conformational strain, exploiting crystallographic occupancies from new refinement techniques (tested in 2.A). B. Modeling ordered water by exploiting Inhomogenous Solvation Theory tested in (tested in 2.B) and also against a more complicated drug target, the μ-opioid GPCR. C. Exploring Higher-levels of ligand partial charges and solvation energies to improve the balance between apolar, polar and ionic energies in docking (tested in 2.C). D. A covalent docking method, seeking adduct formation between electrophilic ligand and a protein nucleophile (tested in 2.D). 2. Testing the new methods in model systems. A. Receptor flexibility is tested in prospective docking screens against a cavity in Cytochrome C Peroxidase (CCP) that bares a flexible loop whose multiple conformations may be observed and weighted from an apo-structure. B. The IST method for Ordered waters is tested against the six waters observed in cavity W191G and the eight in W191G/D192-3, investigating our ability to prospectively predict new ligands that exploit these waters. This work is extended to a more complicated target, the μ-opioid receptor. C. The balance among polar, apolar, and ionic terms is tested vs pairs of cavities differing by single polar substitutions (lysozyme L99A--> L99A/M102H; CCP W191G-->W191G/D235N). D. Covalent docking is prospectively tested against two nucleophilic enzymes, β-lactamase and RSK2 kinase, combining enzymology and x-ray crystallography to evaluate the binding of reversible covalent inhibitors predicted by the new method.