Supporting drug design using an incremental learning approach

Abstract

An intelligence computer system must have the capability to learn. Developing a design support system that is capable of learning is difficult as our understanding of design as an intelligent behaviour is limited and there is, as yet, no computational theory of the design process. From a learning point of view, design can be modelled as an incremental process of discovering the structure of a design problem. Here we present an investigation into how the capability of an AI-based design support system can be enhanced by utilising inductive learning techniques to do this. In this paper, an incremental inductive learning approach to small-molecule drug design is presented. Detailed descriptions of drug design concepts, knowledge representation of drug design objects, and the learning algorithm used are presented to demonstrate how an incremental learning strategy can be integrated into a drug design support system, to provide better solutions to one of the major problems in the domain of small-molecule drug design, the identification of a pharmacophore from a set of active molecules.