Electrophilicity-based charge transfer for developing aquatic-quantitative structure toxicity relationships (Aqua-QSTR)

In this study, aquatic-quantitative structure toxicity relationship (Aqua-QSTR) models for predicting chemical toxicity in aquatic organisms is developed using the electrophilicity-based charge transfer (ECT) descriptor. Firstly, the nature of charge transfer between the selected series of chemical compounds and deoxyribonucleic acid (DNA) bases is carried out to know their electron-donating or accepting nature. Based on the nature of the interaction, Aqua-QSTR studies were carried out for Fathead minnow and Tetrahymena pyriformis using linear regression (LR), machine learning-based random forest regression (RFR), and support vector regression (SVR) methods. LR-derived QSTR on the first set of compounds against Fathead minnow based on maximum ECT values provides 87.7% variation in data with root mean square error (RMSE) of 0.145. Similarly, LR-derived Aqua-QSTR studies on the second set of compounds against Tetrahymena pyriformis based on maximum ECT values give a 90.6% variation in data with an RMSE of 0.163. Further, RFR (SVR) model provides 96.8% (87.7%) variation in data with RMSE of 0.074 (0.145) for the Fathead minnow and 98.1% (90.4%) variation in data with RMSE of 0.073 (0.165) for Tetrahymena pyriformis . The results revealed the utility of ECT in the toxicity prediction of chemical compounds in aquatic organisms.