Nigel Greene, Philip N. Judson, Jan J. Langowski and Carol A. Marchant

Expert Computer Systems for Toxicity and Metabolism Prediction

LHASA UK Ltd., School of Chemistry, University of Leeds, Leeds LS2 9JT, United Kingdom.


It has long been recognized that the ability to predict the metabolic fate of a chemical substance and the potential toxicity of either the parent compound or its metabolites are important aspects in novel drug design. It is also a regulatory requirement that all new chemicals undergo toxicological testing. However, there is increasing pressure to move towards high-throughput screening as well as reducing costs and animal use. The popularity of using computer models to prescreen compounds has grown considerably in recent years.


DEREK, is an expert system designed to identify potential toxicological hazards from chemical structure. It uses rules to describe the known links between structural features of molecules and their toxicity. These rules are developed by an expert panel of toxicologists as part of an on-going collaborative effort. The DEREK rule base covers a wide variety of toxicological endpoints including mutagenicity, carcinogenicity and skin sensitisation.

As part of the refinement process, the predictions for a novel set of compounds are compared to their biological assay results as a measure of the systems performance. For example, 266 non-congeneric chemicals from the National Toxicology Program (NTP) database have been processed through the DEREK mutagenicity knowledge base and the predictions compared to their Salmonella typhimurium mutagenicity data. Initially, 81 of 114 mutagens (71%) and 117 of 152 non-mutagens (77%) were correctly identified. Following further rule base development, the number of correctly identified mutagens has increased to 96 (84%). Further work on improving the predictive capability of DEREK is in progress.


The problems faced by scientists when making risk assessments for existing and novel chemical compounds are enormous. It is often the case that these assessments are based upon incomplete or poor quality data. In these circumstances it is clearly inappropriate to use quantitative methods.

As part of a collaborative project with the Imperial Cancer Research Fund, Logic Programming Associates and City University, LHASA UK has conducted research into new reasoning techniques to address the situations where decisions are required based on data of variable quality. A reasoning model was developed based on the Logic of Argumentation which overcame the difficulty of situations where data could not be represented numerically and hence prevented the application of conventional probability theory.

The model works by constructing the arguments for and against a hypothesis, giving a rigorous assessment of the available data. The aim of the initial project was to build a chemical carcinogenicity advisor with this new reasoning engine forming the core of the technology. The project resulted in a working prototype which independent studies showed to be successful in providing useful advice for a limited range of chemicals. LHASA UK is continuing to develop the prototype into a more commercial product.


Predicting the toxicity of a compound is only one part of the risk assessment process. Chemicals which appear to be benign in themselves may easily metabolize to moieties that are potentially toxic to a biological system. However, it is worth noting knowledge-based systems such as DEREK have metabolism implicit within the rules.

In 1997, LHASA UK embarked on a 3 year project to predict the metabolic fate of novel chemicals. This project, METEOR, will take the new technologies developed for the StAR program and extend and apply them to the prediction of metabolism. The system will use a knowledge-base of biotransformations developed during the course of the project to provide advice on the most likely metabolites.

Keywords: Knowledge-based expert system; toxicity prediction; risk assessment; metabolism prediction; DEREK; StAR; METEOR.
Back to Program Page
Back to Main Page