ABSTRACT
Biological properties of chemical compounds depend in a very sensitive way on their chemical structure. The understanding of such dependencies in terms of socalled structure-activity relationships is one of the principal goals of medicinal chemistry and at the heart of modern computer-aided drug design (CADD). CADD methods have become an indispensable tool in the search for new drugs and may not only aid to guide experimental work effectively but also to elucidate mechanisms of action at the molecular level. Because of the complexity of biological matter and the huge number of chemical structures and possible variations, a large number of quite different CADD methods serving different purposes have been developed. These methods can be roughly categorized into (i) statistical approaches relating physicochemical parameters to biological potency, (ii) heuristic methods usually based on substructural consideration, and (iii) molecular modelling with powerful interactive computer graphics as key instrument. Although the recent development has very much stressed molecular modelling the other methods are still very useful in many cases. This presentation will concentrate on the first category, where so-called quantitative structure-activity relationships (QSARs) are derived with biological potency as dependent and physicochemical parameters as independent. These characterize hydrophobic, electronic and steric properties of drugs and are usually applied to characterize variations of substituents. The most important parameters and their physical meaning will be discussed, together with specific aspects of their application to biological problems. With typical examples from Hansch analysis important aspects of QSARs such as interpretability, predictive power, general QSARs know-how and limitations of QSARs approaches will be outlined. Alternative statistical methods such as classification methods or principal component analysis will also be briefly introduced.
