Prediction of the aqueous solubility of diverse compounds by 2D-QSPR

Abstract Non conformational QSPR models were built for the aqueous solubility (mol/L) at 25 °C of 5610 structurally heterogeneous compounds, including pesticides, drugs and solvents, among others, with a wide range of aqueous solubility. The experimental database was taken from the WATERNT module for solubility estimation of the EPI Suite program. A large set of 23,723 conformation-independent molecular descriptors was explored to obtain trustworthy but simple multivariable linear regression models. It is found that four selected molecular descriptors improve the predictive capacity of the WATERNT program, thus achieving satisfactory predictions for the water solubility through an alternative methodology.