Novel Mathematical Programming Model for Computer Aided Molecular Design

Abstract

The synthesis of new compounds by computer aided molecular design approaches has received considerable attention recently due mainly to the promise of reducing the time and effort required using traditional empirical approaches. These conventional approaches have involved trial and error methods in which a large number of compounds are synthesized and tested in a laboratory. This paper describes a general mathematical programming model for designing compounds that have prespecified performance characteristics. This formulation is novel in that it gives nearly complete information about the molecular structure and is therefore able to exploit accurate property prediction methods that require such information. The computer aided molecular design strategy is able to generate a set of promising candidate compounds for experimental evaluation. The new formulation is illustrated with a refrigerant design case study.