Molecular Property Clustering Techniques

Abstract

In this work, systematic property clustering methods for molecular design are described. Group contribution first-order estimation methods are coupled with property clustering methods to define the targets of the property search space and to predict the properties of the synthesized formulations. In the visual property clustering approach, the property constraints on the molecular synthesis problem are transformed into a feasibility region on a ternary property clustering diagram. Molecular building blocks are represented as points on the diagram, and the synthesis of candidate formulations is achieved by adding molecular fragments using linear mixing rules. Candidate formulations must satisfy necessary and sufficient conditions. The visual clustering methods are limited to problems that can be defined by three properties. For all other cases, the algebraic molecular clustering approach is capable of lowering the complexity of the molecular design problem. Proof-of-concept examples are used to demonstrate the tools.