期刊名称：Journal of Materials and Environmental Science
印刷版ISSN：2028-2508
出版年度：2017
卷号：8
期号：2
页码：476-484
出版社：University of Mohammed Premier Oujda
摘要：Quantitative structure–toxicity relationship (QSTR) models are useful to understand howchemical structure relates to the toxicity of natural and synthetic chemicals. The chemicalstructures of 70 heterogeneous phenols have been characterized by electronic and physicchemicaldescriptors. Density functional theory (DFT) with Beck’s three parameterhybrid functional using the LYP correlation functional (B3LYP/6-31G(d)) calculationshave been carried out in order to get insights into the structure chemical and propertyinformation for the study compounds. The present study was performed using principalcomponent analysis (PCA) method, multiple linear regression method (MLR), multiplenon-linear regression (MNLR) and artificial neural network (ANN). The quantitativemodel of the toxicity of these compounds was accordingly proposed and interpreted basedon the multivariate statistical analysis. The statistical quality of the MLR and MNLRmodels was found to be efficient for the predicting of the toxicity, but when compared tothe obtained results by ANN model, we realized that the predictions achieved by thislatter one were more effective. This model provided statistically significant results andshowed good internal stability and powerful predictability. The squared correlationcoefficients were 0.801, 0.802 and 0.824 for MLR, MNLR and ANN models respectively.The obtained results suggested that the proposed descriptors could be useful to predict thetoxicity of heterogeneous phenols to Tetrahymena pyriformis.
