The use of neural networks in QSARs for acute aquatic toxicological endpoints

Abstract This review surveys the applications of neural network (NN) methodologies to the field of Quantitative Structure–Activity Relationships (QSARs) in aquatic toxicology. Several NN methods have been applied to substantial data sets (some involving over 1000 chemicals) for acute and sublethal toxicity endpoints for fish, invertebrate, protozoan and bacterial species. The results clearly demonstrate the methods' general ability to detect and apply non-linear structure–activity relationships for the prediction of the corresponding values for compounds not part of the training sets.