Machine learning–based QSAR for safety evaluation of environmental chemicals

Safety evaluation of chemicals to human and environmental health is of significance for regulatory decisions. The current high-throughput assays result in a rapid increase of big data for various chemicals, facilitating machine learning (ML)-based quantitative structure–activity relationships (QSAR) modeling. At present, a large number of ML-based prediction models have been innovatively constructed and applied to safety evaluation of various chemicals. Herein, we provide a detailed overview of the development of ML-based QSAR models, especially the model features and relevant algorithms. The databases and predictive schema that can be used for ML-based QSAR modeling are briefly summarized. The current challenges are discussed, and the future developing trends and extended applications of ML-based QSAR modeling in chemical safety evaluation are prospected.