Including toxicity and market availability in a Computer-Aided Molecular Design methodology

Abstract

Advances in the understanding of natural phenomena and the exponential increase of computing power over the last years have made possible the solution of chemical product design problems using computational approaches. In this work, a Computer Aided Molecular Design (CAMD) methodology is proposed and implemented for the design of environmentfriendly solvents in liquid-liquid extraction. As the proposed methodology aims to solve a problem of chemical industry, to ensure that the designed solvents can be easily acquired or synthetized, market availability criteria are included. The proposed CAMD methodology formulates and solves a multi-objective optimization problem where the decision variables are molecules represented as chemical graphs. In the definition of this problem, a first objective is the maximization of solvent power and a second objective is the minimization of environmental impact. Market availability is included in the methodology as one constraint of the optimization problem. In optimization, molecules require specific encodings and the usage of flexible methods. Hence, in the methodology proposed, the evolutionary algorithm HAEA is selected to perform optimization as this algorithm allows flexibility in the representation of individuals and the inclusion of custom genetic operators. The original HAEA is intended to solve single-objective optimization problems, then this work proposes and implements a multiobjective version of HAEA (MOHAEA) for the solution of the optimization problem contained in the CAMD methodology. In MOHAEA, Pareto optimality and the NSGA-II crowding-distance are used to evaluate solutions and guide the evolution. In addition, a strategy for the handling of constraints based on Pareto front punishment is proposed in this new algorithm. The methodology presented in this document is an extension of the CAMD methodology presented by Serrato in 2009. The Serrato's work is the starting point of this work and many of the methods persist in the methodology proposed. Serrato addresses the chemical product design problem of designing optimal solvents for liquid-liquid extraction using a single-objective optimization approach. The study case in both works is the design of optimal solvents for the separation of lactic acid from an aqueous solution and the major improvement of the new methodology proposed is the design of solvents with similar solvent power, a significant reduction of environmental impact and a market availability greater than 80%.