Evolutionary design of molecules with desired properties using the genetic algorithm

Designing new molecules possessing desired properties is an important and difficult problem in the chemical, material, and pharmaceutical industries. The standard approach to this problem consists of an iterative formulation, synthesis, and evaluation cycle that is long, time-consuming, and expensive. Current computer-aided design approaches include heuristic and exhaustive searches, mathematical programming, and knowledge-based systems methods. While all these methods have a certain degree of appeal, they suffer from drawbacks in handling combinatorially large, nonlinear search spaces. Recently, a genetic algorithm-based approach was shown to be quite promising in handling these difficulties. In this paper, we investigate the performance of the basic genetic design framework for larger search spaces. We also present an extension to the basic genetic design framework by incorporating higher-level chemical knowledge to handle constraints such as chemical feasibility, stability, and complexity better. These advances are demonstrated with the aid of a polymer design case study.