Detalle de: Multi-objective evolutionary algorithm for evaluation of shape and electrostatic similarity

Information in chemistry field is increasing each year implying new databases and more available information. Different techniques are adopted to learn how to manage that information. Ligand-Based Virtual Screening methods help in this process. It consists on processing a compound against large databases containing up to millions of chemical compounds evaluating one or more properties. After screening, compounds with the most similar descriptors to those of the target compound are chosen for in-vitro analysis. There exist a large number of molecular descriptors to compare molecules, and in literature, they usually analyze them individually. Regarding that, in this work in progress, we propose a multi-objective algorithm where two descriptors are considered, shape and electrostatic potential similarity. Using these two objective functions, the new algorithm aims to achieve an optimal Pareto-front that allows expert eye to select the most suitable compounds according to the properties of the target compound. Different techniques have been developed to assure fast performance keeping high-quality results. The new algorithm has been compared with different algorithms from the state-of-the-art to evaluate the quality of its results using well-known databases.