Rank-based ant system method for non-linear QSPR analysis: QSPR studies of the solubility parameter

The solubility parameter (δ) plays a unique role in the development of stable pharmaceutical formulations for assessing phase segregation during product synthesis. Understanding this parameter helps to determine how a drug substance will behave when processed or when dosed in vivo. The aim of this work was to develop a novel comprehensive yet rapid and accurate Quantitative Structure-Property Relationship (QSPR) method based on the rank-based ant system feature selection. The method was coupled with the multiple linear regression and support vector regression and applied to the assessment of solubility parameters for a diverse dataset of 1804 chemical compounds. The models were validated by solubility prediction of 360 test set compounds which were not used in building models. The developed models have high prediction power characterized by r (2) values 0.75 and 0.82, and RMSE values 1.96 and 1.65 (J/(cm(3)))(0.5) for the external test set. Various validation techniques and comparison results with the novel optimized support vector regression indicate that the developed models can be used to determine the solubility parameters for a diverse set of chemicals with an acceptable accuracy. The developed models can be beneficial for designing new chemical materials with desired solubility parameter values.