The Zuse Institute Berlin (ZIB) is specialized on efficient algorithms for a model driven analysis of large data sets. We have already established a predictive model for the estimation of the expected (in-vitro) estrogenic effect of trace pollutants based on molecular simulation data. We will combine this model with a method for predicting sites of cytochrome P450-mediated hydroxylation in order to extend our estrogene models to the in-vivo situation. In the long run we want to build computer-based models in order to predict special in-vivo effects of small molecules. At least we want to help to prioratize animal testings according to computer-based predictions.