Replacing affinity with binding kinetics in QSAR studies resolves otherwise confounded effects

Steady-state binding affinity is a commonly used response variable in quantitative structure-activity relationship QSAR) studies. In this paper we show that affinity alone may introduce unnecessary noise to the model or even cause model failure. Binding affinity is a ratio between the dissociation rate and association rate. Mathematical modeling of a ratio with unknown variation structure of nominator and denominator is difficult and the interpretation of the resulting model might be meaningless under certain circumstances. Commercially available biosensors can be used for measurement of binding kinetics enabling separate modeling of association and dissociation rates. In this paper we discuss the problems caused by the confounded effect in the compiled response affinity and show examples where separate modeling of the rate constants improves the interpretation of QSAR models. Copyright (c) 2007 John Wiley & Sons, Ltd.