Application of principal component-genetic algorithm-artificial neural network for prediction acidity constant of various nitrogen-containing compounds in water

摘要

Abstract Principal component-genetic algorithm-multiparameter linear regression (PC-GA-MLR) and principal component-genetic algorithm-artificial neural network (PC-GA-ANN) models were applied for prediction acidity constant (pK _a ) for various nitrogen-containing compounds. A data set that consisted of 282 various compounds, including 55 anilines, 77 amines, 82 pyridines, 14 pyrimidines, 26 imidazoles and benzimidazoles, and 28 quinolines, is used in this work. A large number of theoretical descriptors were calculated for each compound. The first 179 principal components (PCs) were found to explain more than 99.9% of variances in the original data matrix. From the pool of these PCs, the genetic algorithm was employed for selection of the best set of extracted PCs for PC-MLR and PC-ANN models. The models were generated using 15 PCs as variables. For evaluation of the predictive power of the models, pK _a values of 56 compounds in the prediction set were calculated. Root mean square errors (RMSE) for PC-GA-MLR and PC-GA-ANN models are 1.4863 and 0.0750. Comparison of the results obtained by the models reveals superiority of the PC-GA-ANN model relative to the PC-GA-MLR model. Mean percent deviation for the PC-GA-ANN model in the prediction set is 2.123. The improvements are due to the fact that pK _a of the compounds demonstrates non-linear correlations with the PCs.