Modeling of the total elimination half life for environmental chemicals

Quantitative Structure Activity Relationship (QSAR) analysis was implemented for the prediction of elimination half-life for environmental chemicals. The experimental values of elimination half-life and two sets of molecular descriptors, the Linear Free Energy Relationship (LFER) and the PaDEL descriptors, were collected for 199 environmental chemicals and used as input data for the development of QSAR models. The initial datasets were split to the training and the prediction set. Principal Component Analysis (PCA) was implemented for the distribution of chemical compounds, while a genetic algorithm was then used for the selection of the optimal set of descriptors. The datasets were analyzed using Multiple Linear Regression (MLR) and Artificial Neural Networks (ANN). The fitting performance (R2) of the best models, using LFER and PaDEL descriptors, was 0.80 and 0.87, respectively. The Applicability Domain (AD) of the developed models was determined indicating that there were no outliers and verifying their reliability. The developed QSARs were then applied to several compounds with unknown values of elimination half-life. In conclusion, the proposed models were successfully evaluated for their fitting capacity, their validity and applicability and were found to be capable of predicting the elimination half-life of “data poor” chemical compounds within their applicability domain.