Identifying Organic\nChemicals with Acetylcholinesterase\nInhibition in Nationwide Estuarine Waters by Machine Learning-Assisted\nMass Spectrometric Screening

Neurotoxicity is frequently observed in the global aquatic\nenvironment,\nthreatening aquatic ecosystems and human health. However, a very limited\nproportion of neurotoxic effects (∼1%) has been explained by\nknown chemicals of concern. Here, we integrated machine learning,\nnontargeted analysis, and in vitro biotesting to identify neurotoxic\ndrivers of acetylcholinesterase (AChE) inhibition in estuarine waters\nalong the coast of China. Machine learning was used to predict AChE\ninhibitors in a large chemical space. The prediction output was profiled\ninto a suspect screening list to guide high-resolution mass spectrometry\n(HRMS) screening of AChE inhibitors in estuarine water samples. Ultimately,\n60 chemicals with diverse known and presently unknown structures were\nidentified, explaining 82.1% of the observed AChE inhibition. Polyunsaturated\nfatty acids were unexpectedly found to be neurotoxic drivers, accounting\nfor 80.5% of the overall effect. This proof-of-concept study demonstrates\nthat machine learning-based toxicological prediction can achieve a\nvirtual fractionation role to pinpoint HRMS features with the bioactivity\npotential. Our approach is expected to enable rapid and comprehensive\nscreening of organic pollutants associated with various in vitro end\npoints for large-scale monitoring of water quality.