A COMPARATIVE-STUDY OF MOLECULAR SIMILARITY, STATISTICAL, AND NEURAL METHODS FOR PREDICTING TOXIC MODES OF ACTION

Quantitative structure-activity relationship (QSAR) models are routine ly used in predicting toxicologic and ecotoxicologic effects of untest ed chemicals. One critical factor in QSAR-based risk assessment is the proper assignment of a chemical to a mode of action and associated QS AR. In this paper, we used molecular similarity neural networks, and d iscriminant analysis methods to predict acute toxic modes of action fo r a set of 283 chemicals. The majority of these molecules had been pre viously determined through toxicodynamic studies in fish to be narcoti cs (two classes), electrophiles/proelectrophiles. uncouplers of oxidat ive phosphorylation. acetylcholinesterase inhibitors, and neurotoxican ts. Nonempirical parameters, such as topological indices and atom pair s, were used as structural descriptors for the development of similari ty-based, statistical, and neural network models. Rates of correct cla ssification ranged from 65 to 95% for these 283 chemicals.