|
1F - QSAR (TUP/42) Computational Prediction Models for Aquatic Toxicity. Netzeva, Tatiana1, Benfenati, Emilio2, Lo Piparo, Elena2, Gini, Giuseppina3, Schultz, Terry4, Cronin, Mark 1, 1 Liverpool John Moores University, Liverpool, Merseyside, England2 Istituto di Ricerche Farmacologiche “Mario Negri”, Milan, Milan, Italy3 Politecnico di Milano, Milan, Milan, Italy4 The University of Tennessee, Knoxville, TN, USA ABSTRACT- Proposed changes in European legislation for registration, evaluation and authorisation of new and existing chemical substances will require the development of rapid, reliable and validated alternative methods to generate toxicological and ecotoxicological information. Computational prediction models (PM) undoubtedly have a key role in the process of reduction and replacement of animal testing for regulatory purposes. Key amongst these is the use of quantitative structure-activity relationships (QSARs) and extrapolations of toxicity from one species to another. The aim of this study was to collect high quality ecotoxicity data for the purpose of the development of PMs. Two data sets were identified, namely those containing toxicity data to the fish Pimephales promelas and the ciliate Tetrahymena pyriformis. A database of 970 organic compounds with acute toxicity data to T. pyriformis and/or P. promelas was collected. 375 of the compounds were common to both species. QSARs are reported for individual endpoints and for the inter-species relationships where the inclusion of physico-chemical parameters greatly assisted in the prediction of toxicity. In addition to multiple linear regression, partial least squares and neural networks were applied to the development of prediction models. [This work was supported in part by the European Union IMAGETOX Research Training Network (HPRN-CT- 1999-00015)]. Key words: aquatic toxicity, QSAR, interspecies relationship, fish |
