Better prediction of aqueous solubility of chlorinated hydrocarbons using support vector machine modeling

Remediation of water contaminated by organic pollutants is a major challenge, which could be improved by better knowledge on the aqueous solubility of organic compounds. Indeed, the aqueous solubility controls the fate and toxicity of pollutants. Here we performed a structure–property study based on a genetic algorithm for the prediction of aqueous solubility of chlorinated hydrocarbons. 1497 descriptors were calculated with the Dragon software. The variable selection method of the genetic algorithm was used to select an optimal subset of descriptors that have significant contribution to the overall aqueous solubility, from the large pool of calculated descriptors. The support vector machine was then employed to model the possible quantitative relationships between selected descriptors and aqueous solubility. Our results show that total size, polarizability and electronegativity modify the aqueous solubility of compounds. We also found that the support vector machine method gave better results than other methods such as principal component regression and partial least squares.