Predicting acute toxicity from a process-based perspective

Tjalling Jager, Helene Magaud, Erik van de Plassche, Bas Kooijman

Estimating acute LC50s from hydrophobicity is one of the earliest successful QSAR 
applications. However, the LC50 has limitations as a measure of toxicity; LC50s decrease in 
time in a manner that depends both on the toxicant and the species tested. For hydrophobic 
compounds and large fish, the standard test duration may be insufficient to reach steady state 
in the organism. Therefore, the resulting LC50 may not represent the ultimate value, and 
thereby bias is introduced in QSARs. To develop better prediction methods, it is advisable to 
focus on parameters with a closer link to the underlying processes. We explore the 
possibilities of process-based models in QSAR development by applying the DEBtox method 
to acute toxicity data for fathead minnows. DEBtox uses all of the observations in time to fit 
three time-independent parameters, which are closely linked to the underlying processes and 
provide a better starting point for identifying patterns in the data. For narcotic toxicants, these 
parameters are related in a predictable manner to each other and to hydrophobicity, making it 
easier to identify compounds with a specific mode of action.