MechoA+:A Chemical Structure Profiler Raising theBar for the Prediction of Mechanisms of Toxic Action for ChemicalSafety Assessment

With stakeholders in chemical regulation increasingly advocating for nonanimal testing methodologies, there is a need for reliable <i>in silico</i> tools with wide applicability domains to predict both environmental and human hazards. In this context, the <i>in silico</i> structure-based MechoA+ scheme has been developed to predict molecular initiating events, appropriate for both mammalian and ecotoxicity by merging and refining two previous classification models, MechoA and Sapounidou–Firman schemes. The resulting model is a new decision tree composed of 152 structural alerts able to classify a wide range of substances within 6 mechanistic classes and 27 subclasses and two rules excluding substances out of the scope of the scheme. Analysis of MechoA+ scheme predictions shows a higher percentage of valid predictions (92% predicted positive value on the training set) and wider structural, mechanistic, and taxonomic domains than the previous models on a data set of more than 70,000 substances (covering cosmetics, pesticides, etc.), achieving predictions for 80% of substances. Since MechoA+ is implemented within readily available software tools, its widespread adoption will facilitate more accurate hazard assessments, QSAR building, read-across, and grouping, strengthen regulatory decision-making, and support safer chemical design in early-stage research and development.