Chapter 11. Expert Systems: The Use of Expert Systems in Drug Design-Toxicity and Metabolism

Abstract

In this chapter, we provide an overview of some of the characteristic features of knowledge-based expert systems for the prediction of toxicity and xenobiotic metabolism, and highlight the application of some of their uses in drug discovery and development, including the increasingly important area of risk assessment and regulatory submissions. A discussion of structure–activity relationship applicability domains is included and, in the context of toxicology prediction, we foreground the use of the expert system approach in the assessment of genotoxic impurities. We revisit some previously reviewed functionality of expert systems and illustrate how numerical data and quantitative estimations of molecular properties can be used by rules in the expert systems’ in-built reasoning engines to attenuate the force of predictions made by the system. We compare expert systems to quantitative tools but highlight the complementarity in the natures of quantitative and qualitative approaches and illustrate how their concerted use can increase predictivity to a level higher than may be attainable by individual methods in isolation. In the context of metabolism prediction, we indicate how a quantitative method that predicts sites of cytochrome P450-mediated metabolism of drug-like molecules can be use to refine predictions offered by the expert system. We also explain how the expert system metabolism predictor can in assist in structure elucidation and characterisation of metabolites through the use of data acquired by mass spectrometric analysis. We outline some approaches to the evaluation of expert systems and how these, out of necessity, differ from validation methods as applied to quantitative methods. We further discuss some of the benefits of combining different expert systems within the framework of higher level reasoning protocols.