An alternative to in vivo/in vitro toxicity testing: Consensus-based prediction for toxicity assessment and prioritisation of diverse chemicals in Japanese eel to safeguard aquatic and public health.

Abstract

Fish is widely consumed as a rich source of protein that provides substantial nutritional and therapeutic benefits to the human population. Japanese eel is one of the most consumed fish in Japan, China, and other Asian countries, as a rich source of protein and omega-3 fatty acids. Unfortunately, it is classified as an endangered species by the International Union for Conservation of Nature (IUCN). This is due to the deposition of harmful chemicals into the aquatic environment via improper sewage disposal, agricultural runoff, and industrial discharge, subsequently deteriorating water quality. This causes serious threat to the life of aquatic organisms, consequently affecting the human beings. In this research, the first-ever quantitative structure-activity relationship (QSAR) models are developed to predict the aquatic toxicity of diverse industrial chemicals towards Anguilla japonica, encompassing the endpoint median lethal concentration (LC50). Intelligent consensus prediction (ICP) was implemented to overcome the estimation bias of individual models for a specific query compound, by providing greater coverage of chemical space. The statistical metrics obtained suggest that the developed QSAR models are good-fit, robust, and predictive. Consensus modeling outperformed the individual models in terms of the external validation metrics (Q2F1 = 0.843, Q2F2 = 0.840 and Q2F3 = 0.833) and mean absolute error of the test set (MAEtest(95%) = 0.417) compounds. The potential toxicants were identified by screening the pesticides properties database (PPDB) to check the robustness and reliability of the models. The specific biomarkers responsible for toxicity were identified. This aids in designing safe and green chemicals, supporting greater ecosystem diversity.