QSPR Study on the Soil-Water Partition Coefficient of Polychlorinated Biphenyls by Using Artificial Neural Network

Abstract

A practicable quantitative structure property relationship (QSPR) model for predicting the soil-water partition coefficient, Koc, of 16 polychlorinated biphenyls (PCBs) was developed. The structure of the investigated PCBs is encoded by five quantum structural descriptors and on topological index. The calibration model of Koc was developed by using artificial neural network (ANN). The input variables of ANN were generated from 6 structural descriptors by using principal component analysis (PCA). Leave one out cross validation was carried out to assess the predictive ability of the developed model. The prediction RMS%RE for the 16 PCBs is 6.35. The R2 between the predicted and experimental logKoc is 0.8522. It is demonstrated that ANN combined with PCA is a practicable method for developing QSPR model for Koc of these PCBs.