Quantitative Structure Toxicity Relationship (QSTR) Models for Predicting Toxicity of Polychlorinated Biphenyls (PCBs) Using Quantum Chemical Descriptors
Sabitu Babatunde Olasupo, Adamu Uzairu, Balarabe Sarki Sagagi · International Journal of Bioorganic Chemistry · 2017
The Density functional theory (DFT) at B3LYP of 6-31G* basis set was employed to optimize 30 polychlorinated Biphenyls (PCBs) involved in this study by using Genetic function appropriation algorithm (GFA) approach to develop regression models in order to predict the toxicity of the compounds. The optimum model which has squared correlation coefficient (R2) = 0.9382, cross validated correlation coefficient (R2cv) = 0.9056, adjusted squared correlation coefficient (R2Adj) = 0.9228 and external prediction (R2pred) =0.7238 was selected. The robustness of the model was confirmed by method of Y- randomization and the accuracy of the proposed model was also illustrated by using cross-Validation, validation through an external test set and applicability domain techniques. This QSTR model proved to be a useful tool in the prediction of toxicity of the congeneric compounds and a guide in the identification of structural features that could be responsible for toxicity of other polychlorinated aromatic compounds.