Knowledge‐Based Approaches for Predicting the Sites and Products of Metabolism
Philip Neville Judson · Methods and principles in medicinal chemistry · 2014
Knowledge-based expert systems (KBS) are computer artificial intelligence systems that use human knowledge to reason about problems or to make predictions. They can make qualitative judgements on the basis of qualitative evidence whereas, to perform well, statistical and other numerical methods depend on quantitative or categorical evidence. KBS to predict potential chemical toxicity and to predict xenobiotic metabolism have been available for over twenty years and are now in routine use. Predicting metabolism is challenging: it is fairly easy to design a system to predict all potential metabolites; the challenge is to restrict predictions to those metabolites that will be found in laboratory studies. Reasoning methods to limit predictions to the metabolites most likely to be observed are described. Finally, there is a short discussion about problems with assessing predictive accuracy of KBS, some of which are peculiar to metabolism prediction.