Tutorial 7 www.moleculardescriptors.eu

QSARs establish a quantitative relationship between chemical structures and their properties [1]. In theory, QSAR models can be used to predict the properties of chemical structures, provided their structural information is available. In the recent years, there had been a growing awareness about QSARs and their applications. This is quite evident from their use for regulatory purposes. A new European legislation on chemicals – REACH (Registration, Evaluation, Authorization and restriction of Chemicals) came into force in 2007, allows and encourages the use of QSAR model predictions when the experimental data are not sufficiently available or as supplementary information, provided validity of the model is justified [2,3]. However, this rising popularity of QSAR models is also accompanied by a question over their reliable predictions [4]. In theory, derivation of QSAR models is based primarily on training sets which are structurally limited and thus, their applicability to the query chemicals is limited [5]. Thus, their applicability towards reliable predictions is restricted in a chemical space to some specific chemical categories. Such reliable predictions are usually confined to those chemicals, that ar