It is an important procedure to evaluate chemical aquatic toxicity for facilitating water resource management and environmental protection. Compared with experimental methods, computational methods can alleviate the dependence on animal experiments and reduce the cost of time and money, thereby being more suitable for extensive evaluation of chemicals. However, most of the current aquatic toxicity prediction models fail to simultaneously meet the demands for ecological risk assessment and molecular design, due to the limited applicability domain or unsatisfactory performance. Therefore, we developed a set of probabilistic classification models for predicting aquatic toxicity using ensemble learning algorithm. The chemical toxicity in five aquatic species were studied. The ensemble models was stacked based on six single models to make the best of experimental toxicity datasets encoded with structure graphs and molecular descriptors. As a result, the ensemble models universally outperform single models, achieving areas under the curves (AUC) of 0.75~0.92 and F1 scores of 0.65~0.87. Additionally, a knowledge base of aquatic toxicity modes of action (MOA) was established for toxicity mechanism analysis. Further, a server integrated with these ensemble models and the knowledge base—AquaticTox (http://chemyang.ccnu.edu.cn/ccb/server/AquaticTox/) was constructed, providing a platform to rapidly classify aquatic toxicity and assign MOA categories for chemicals. Having applied in pesticide molecular design, it is demonstrated that our ensemble models and platform effectively facilitate the prioritization of chemical toxicity testing and the design of eco-friendly molecules.

AquaticTox: An ensemble learning method for predicting chemical aquatic toxicity

Where thousands of chemistry professionals meet to share ideas and advance scientific and technical knowledge. Spring 2023 abstracts, posters, and presentations will launch at the end of March.

ACS Spring 2023

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.