Neural Network Based QSPR Study for Predicting pKa of Phenols in Different Solvents

Computational Neural Network (CNN) based QSPR methodology is applied to a multicomponent system: the prediction of p K a of phenols in different solvents. The system is composed of 94 phenols, 10 solvents, and 276 experimental p K a values. The phenols are characterized by the habitual molecular descriptors, while the solvents are described by a number of physical properties and by several parameters of the most used multiparametric polarity solvent scales. The proposed model, non‐linearly derived, contains seven descriptors; five of them belong to the solutes and the other two to the solvents. Good results are obtained with a Root‐Mean‐Square Error (RMSE) and correlation coefficients ( R 2 ) of 0.71 (0.982), 0.83 (0.977), and 0.95 (0.975) for the training, prediction, and cross validation sets, respectively. The robustness of the model is also in accord with the statistical results obtained from particular subsets of phenols with and without ortho ‐substituents and those obtained from the subsets of values of p K a determined in protic or in aprotic solvents and also in each solvent. The descriptors of the model encode information that reflects characteristics of the molecules of the solutes and the solvents clearly related to the interactions acting in the dissociation process.