CHAPTER 6. Quantitative Structure–Activity Relationships

Abstract

Quantitative structure–activity relationships (QSARs) are empirical statistical models commonly used by medicinal chemists in compound design. They have had a long history of successful use in the pharmaceutical industry, and developed out of an even longer history of quantitative structure–reactivity relationships in physical organic chemistry. Originally, QSARs were used to model within close homologous series, but they have become more widely useful in modelling drug metabolism and pharmacokinetic properties across project series, which brings new challenges. Assessment of quality of the QSAR model, and its ability to make accurate and precise predictions for a particular chemical series, is a critical question for medicinal chemists. This chapter describes good practice in building and validating QSAR models, and also some of the common language used in building and validating QSAR models, to aid the dialog between medicinal chemists and computational chemists. QSAR model building can appear to medicinal chemists as a collection of intimidating statistical technologies and hard to interpret molecular descriptors. Visual inspection of structure–activity relationships through graphing physical properties vs activity, often a prelude to QSAR modelling is highlighted as a very powerful technique in itself. Matched molecular pair analysis, an “inverse QSAR” method, is also described, and high impact examples of successful QSARs and useful QSAR tools, freely available on the internet, are highlighted.