Protein-ligand docking using a multi-objective genetic algorithm.

Ligand docking is the computational prediction of the bound conformation of a small molecule in a complex starting from its atomic coordinates. A docking procedure is composed of two contributions: a scoring function determining the relative score of different configurations of the ligand and a search procedure exploring the space of the possible ligand orientations and conformations. Scoring functions are often based on calculation of binding energies between ligand and receptor; the search procedures have to find the ligand pose that this corresponds to the absolute minimum of the binding energy landscape. Binding energies are sums of different contributions. Search procedures generally look for the minimum of the global scoring function neglecting the possibility to optimize the search towards each single member or a group of them (for example bonded and non-bonded energies). This work describes the application of a Multi-Objective Genetic Algorithm (MOGA) as a search procedure for a protein-ligand docking problem. The MOGA implementation in modeFRONTIER™ is used to optimize the docking towards each of the different contributions of the docking program Autodock v. 3.05 (http://autodock.scripps.edu) scoring function. MOGA results give a set ofsolutions, each one representing a compromise between the different objectives (the so called Pareto frontier). This could be useful, for example, to get a docked complex that exalts one of the contributions to the total binding energy. The results consist of a set of docked ligand conformations within the 2-objectives Pareto frontier (minimization of bonded energy and minimization of non-bonded energy) having a root mean squared deviation from the crystal structure smaller than 1.5 Å. From the inspection of the Pareto frontier results that the non-bonded interactions effectively have a greater influence than bonded interactions in guiding the search.