The combination of chromatographic retention and molecular descriptors to predict membrane-passage properties of pharmaceuticals.

A major problem in the development of new pharmaceuticals is that a lot of substances, potentially useful, fail in a later phase, due to not appropriate ADME-Tox properties. Therefore it is necessary to develop methods, which can predict these properties in an early phase of the development.
In this project we will limit us to the prediction of membrane-passage properties of molecules (in extensio to gastro-intestinal absorption and blood-brain-barrier passage). We will use known in-vitro, in silico and chromatographic methods, which are used to predict absorption, to generate parameters. These parameters will be used as descriptors in order to generate QSAR-models for the prediction of absorption.
For the selection of the most significant descriptors, as well as for modeling we will use the CART- and MARS-methods. Two methods which are not yet being used in this area. We will evaluate the performance of these two methods and compare them with more conventional methods.
We will build CART and MARS-models for datasets found in the literature, which we will extend with the results of the measurements of the above mentioned in-vitro and chromatographic methods.
Our goal is to generate a model, based on a limited set of descriptors, which give a general idea of the absorption properties of molecules. Based on this model one should be capable to in silico screen a set of molecules and eliminate those with poor absorption properties in an early phase of the development process.