In silico methods for toxicity prediction

Abstract

The principles and uses of (Q)SAR models and expert systems for predicting toxicity and the biotransformation of foreign chemicals (xenobiotics) are described and illustrated for some key toxicity endpoints, with examples from the published literature. The advantages and disadvantages of the methods and issues concerned with their validation, acceptance and use by regulatory bodies are also discussed. In addition, consideration is given to the potential application of these techniques in regulatory toxicity testing, both individually and as part of a chemically-based read-across approach, particularly for the risk assessment of chemicals within intelligent, integrated decision-tree testing schemes. It is concluded that, while there has been great progress in recent years in the development and application of in silico approaches, there is still much that has to be achieved to enable them to fulfill their potential for regulatory toxicity testing. In particular, there is a need for the wider availability of appropriate biological data and international agreement on how the systems should be validated. In addition, it is important that correlations between activity and physicochemical properties are based on a mechanistic basis to maximize the predictivity of models for novel chemicals.