Review of Quantitative Structure - Activity Relationships for Acute Mammalian Toxicity

This paper reviews Quantitative Structure-Activity Relationship (QSAR) models for acute mammalian toxicity published in the last decade. A number of QSAR models based on cytotoxicity data from mammalian cell lines are also included because of their possible use as a surrogate system for predicting acute toxicity to mammals. On the basis of the review, the following conclusions can be made: i) a relatively small number of models for in vivo toxicity are published in the literature. This is due to the nature of the endpoint - acute systemic toxicity is usually related to whole body phenomena and therefore is very complex. The complexity of the mechanisms involved leads to difficulties in the QSAR modelling; ii) most QSAR models identify hydrophobicity as a parameter of high importance for the modelled toxicity. In addition, many models indicate the role of the electronic and steric effects; iii) most of the literature-based models are restricted to single chemical classes. Models based on more heterogeneous data sets are those incorporated in expert systems. In general, the QSAR models for mammalian toxicity identified in this review are considered useful for investigating the mechanisms of toxicity of defined chemical classes. However, for predictive purposes in the regulatory assessment of chemicals most of the models require additional information to satisfy internationally agreed validation principles. In addition, the development of new models covering larger chemical domains would be useful for the regulatory assessment of chemicals