A simple approach for prediction of Henry's law constant of pesticides, solvents, aromatic hydrocarbons, and persistent pollutants without using complex computer codes and descriptors

A novel approach is introduced for reliable prediction of Henry's law constant of persistent pollutants, pesticides, aromatic hydrocarbons, and solvents, which have extensive use for the description of the movement of chemical compounds inside and outside aquatic ecosystems. The largest available experimental data of Henry's law constant for 530 heterogeneous chemicals are used to develop and test the novel model. This method needs the molecular structure of the chosen heterogeneous compound that is based on four non-additive factors including the contributions of hydrogen bonding functional groups, polar groups, halogenated compounds, hydrocarbons as well as the number of specific atoms as additive parameters. Internal and external validations are done on the estimated results for 353 and 177 chemicals of training and test sets, respectively. Various statistical parameters containing correlation coefficient (R-2), the maximum value of errors (MaxError), mean error (ME), and root mean squared error (RMSE) are also used to confirm the high reliability of the novel correlation as compared with the best existing method, which requires complicated descriptors. The values of training set R-2, MaxError, ME, and RMSE for the new/comparative models are 0.9512/0.8622, 1.521/4.417, 0.0000/0.0194, and 0.4632/0.7797, respectively. The same trend the mentioned statistical parameters also exists for the test set, which confirms the new correlation has higher reliability, goodness-of-fit, accuracy, and precision in comparison to the best available method. (C) 2022 Institution of Chemical Engineers. Published by Elsevier Ltd. All rights reserved.