Prediction of the clearing temperatures of a series of liquid crystals from molecular structure

A quantitative structure-property relationship (QSPR) investigation was per formed to develop a mathematical link between molecular structure and the c learing temperature of a series of structurally related liquid crystals. Mo lecular structures were encoded by a series of numerical descriptors encodi ng information regarding size, shape, and the ability to participate in int ermolecular interactions. A genetic algorithm feature selection routine was utilized to select high-quality subsets of these descriptors for use in co mputational neural network models. A successful 10-descriptor model was dev eloped using 318 compounds with a root-mean-square error of 5.4 K for the c learing temperature for the compounds in an external prediction set not use d in model development.