QSAR and Expert Systems in the Prediction of Biological Activity
Philip Judson · Pesticide Science · 1992
Abstract Traditional quantitative structure‐activity relationship (QSAR) approaches and expert system approaches to the prediction of biological activity on the basis of chemical structure have concentrated on different aspects of the problem. The key aim of QSAR studies has been to create models that can predict activity–in a sense, to discover the rules that govern the relationships between the presence of structural features and biological activity. Expert systems developed in recent years, on the other hand, seek to apply rules that encompass existing human knowledge about the links between key, complex structural features and particular toxicological end‐points. It is now becoming possible to provide automated help with the discovery of rules that are suitable for use in the knowledge‐based expert systems. Some examples of this are discussed.