Chemometric methods
Chemometric methods derive mathematical models (quantitative structure activity relation- ships: QSARs) describing a biological activity of interest (or a profile of such activities) in terms of chemical parameters (measured, from tabulations, or computed). A variety of such methods as, for example, multiple regression analysis, principal component and factor analysis, principal component regression, PLS, cluster analysis, discriminant analysis, SIMCA, support vector machines, neuronal nets, or topology based logical approaches etc. are available at CDD GbR. QSAR models allow to make estimates of biological properties of new compounds and can thus aid in decision making for further syntheses and/or biological testing. They can also provide information on the mechanism of action.
Directed at saving experimental work and increasing the chances of success in drug discovery projects chemometric methods offer indispensable support
for the extraction of information from the ever increasing amount of data
to address pharmacokinetic as well as possible aspects of adverse or toxic effects of therapeutic agents or agrochemicals at an early stage of compound development
for "hit to lead" decisions and lead optimization taking into account results from tests such as Caco-2(PgP), herg, CYP-inhibition etc.
for the design of libraries or representative training series of compounds
   
The thousands of known QSARs attest to the versatility of chemometric drug design methods, and QSARs have been used successfully in many cases of new compound design. Series design increases the information gained per compound tested.