Classification models for aqueous reaction rate constants of hydroxyl radical with organics
摘要
Hydroxyl radicals (•OH) can unselectively and rapidly react with numerous species present in water. The reaction rate constant (kOH) of •OH with organics is vital in evaluating its persistence in aqueous phase. This work, for the first time, reported a global classification model for predicting the categories of logkOH, by using random forest (RF) algorithm. The optimal RF model was based on 11 Dragon descriptors, yielding predictive accuracies of 97.9% (training set of 746 organics), 95.2% (test set of 249 organics) and 97.2% (total data set of 995 organics). The classification model in this work provided an efficient tool for assessing logkOH categories. The investigation suggests that introducing halogen atoms (e.g. Br) to organic compounds would reduce the reaction probability being attacked by •OH radicals. While multiple bonds (e.g., double, triple and aromatic bonds), rotatable bonds, or groups with more hydrogens present in terminal carbon atoms can make a positive contribution to logkOH.
Graphical abstract