On the interpretability of quantitative structure-activity relationships (QSAR).

QSAR are a powerful tool to obtain information on the molecular mechanism of drug action as well as on properties of receptors. They can be evaluated at different levels of sophistication using different parameters to describe physical chemical properties of the drugs and different mathematical methods. In this presentation only QSAR at the level of extrathermodynamic parameters will be considered. Interpretation of QSAR is always connected with two basic questions: 1) Which requirements must be met in order to get meaningful (interpretable) QSAR and what can be done to satisfy these requirements? 2) What information can be gained (and in which way) from QSAR? Some of the most important aspects of these questions to be discussed in this paper are: 1) Interrelationships between parameters and the separability of hydrophobic, electronic and steric effects. 2) Description of hydrophobic interactions. 3) Separation of hydrophobic effects connected with transport and hydrophobic bonding. 4) Estimation of the biological point of attack. 5) Selection and informational content of biological response data. Factor analysis will be shown to be a useful data preprocessing step allowing one to systematize data and to recognize and to eliminate collinearities. Another multivariate method, principal component analysis, may be used to separate pharmacokinetic and pharmacodynamic effects.