Philip N. Judson

Knowledge-Based Expert Systems for Toxicity and Metabolism Prediction: DEREK, StAR and METEOR

It has long been recognised 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 in novel drug design. The popularity of using computer models as an aid in this area has grown considerably in recent years. LHASA Limited has been developing knowledge-based expert systems for toxicity and metabolism prediction in collaboration with industry and regulatory authorities. These systems, DEREK, StAR and METEOR, use rules to describe the relationship between chemical structure and either toxicity in the case of DEREK and StAR, or metabolic fate in the case of METEOR. The rule refinement process for DEREK often involves assessing the predictions for a novel set of compounds and comparing them to their biological assay results as a measure of the systems performance. For example, 266 non-congeneric chemicals from the National Toxicology Program 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 knowledge base development, the number of correctly identified mutagens has increased to 96 (84%). Further work on improving the predictive capabilities of DEREK, StAR and METEOR is in progress.

Using absolute and relative reasoning in the prediction of the potential metabolism of xenobiotics.

To be useful, a system which predicts the metabolic fate of a chemical should predict the more likely metabolites rather than every possibility. Reasoning can be used to prioritize biotransformations, but a real biochemical domain is complex and cannot be fully defined in terms of the likelihood of events. This paper describes the combined use of two models for reasoning under uncertainty in a working system, METEOR-one model deals with absolute reasoning and the second with relative reasoning.

Using argumentation for absolute reasoning about the potential toxicity of chemicals.

The application of a new argumentation model is illustrated by reference to DEREK for Windows, a knowledge-based expert system for the prediction of the toxicity of chemicals. Examples demonstrate various aspects of the model such as the undercutting of arguments, the resolution of multiple arguments about the same proposition, and the propagation of arguments along a chain of reasoning.

Toxicological knowledge discovery by mining emerging patterns from toxicity data

Predicting the risk of toxic and environmental effects of chemical compounds is of great importance to all chemical industries [1]. Expert systems have shown success in predicting toxic risk by applying established knowledge of toxicology encoded as a knowledge base of structural alerts and a reasoning model. A disadvantage of expert systems is that developing new structural alerts requires considerable time and effort from domain experts. In order to expedite this process a software tool has been developed that can automatically mine representations of activating features directly from toxicity datasets and present them in an interpretable form. Our knowledge discovery tool applies emerging pattern (EP) mining [2]: a form of association rule mining [3] that is well known to computer science, but is relatively new to chemistry [4]. The EP mining algorithm accepts any data expressed as a series of binary properties, which is divided into two classes, and extracts patterns of those properties that are frequent within the data and are more frequent in one data class compared to the other. By mining emerging patterns from toxicity datasets, encoded as fingerprints of binary descriptors, the tool generates patterns of features that distinguish toxicants from innocuous compounds. These patterns represent potentially activating features of the toxic compounds that may then be used to define new alerts. The knowledge discovery tool has been tested using a public dataset of 3489 mutagens and 2981 non-mutagens, encoded as fingerprints of approximately 2000 functional groups and ring descriptors. EPs were produced and grouped into a number of hierarchical families. Six of the EPs that represented distinct chemical classes were selected for manual inspection by a toxicology expert. Relevant literature was analysed to find a mechanistic rationale for the mined features, which resulted in four new structural alerts for in vitro mutagenicity.

Assessing confidence in predictions made by knowledge-based systems

A new metric, “veracity”, is proposed for assessing the performance of qualitative, reasoning-based prediction systems that takes into account the ability of these systems to express levels of confidence in their predictions. Veracity is shown to be compatible with concordance and it is hoped that it will provide a useful alternative to concordance and other Cooper statistics for the assessment of reasoning-based systems and for comparing them with other types of prediction system. A few datasets for four end points covered by the program, Derek for Windows, have been used to illustrate calculations of veracity. The levels of confidence expressed by Derek for Windows in these examples are shown to carry meaningful information. The approach provides a way of judging how well open predictions (“nothing to report” in Derek for Windows) can support qualified predictions of inactivity.

Using new reasoning technology in chemical information systems.

Unreliability of numerical data causes difficulties in computer systems for decision-making, risk assessment, and similar activities. Much human judgment is non-numerical and able to make useful evaluations of alternatives under uncertainty. The Logic of Argumentation (LA) offers a basis for computerized support of decision-making in the absence of numerical data, and it is being used in a project on carcinogenic risk assessment, StAR. There are potential applications of LA in other artificial intelligence systems in chemistry, such as for synthesis planning.

Qualitative risk assessment fulfils a need

Classically, risk is characterised by a point value probability indicating the likelihood of occurrence of an adverse effect. However, there are domains where the attainability of objective numerical risk characterisations is increasingly being questioned. This paper reviews the arguments in favour of extending classical techniques of risk assessment to incorporate meaningful qualitative and weak quantitative risk characterisations. A technique in which linguistic uncertainty terms are defined in terms of patterns of argument is then proposed. The technique is demonstrated using a prototype computer-based system for predicting the carcinogenic risk due to novel chemical compounds.

A comprehensive approach to argumentation.

A reasoning model, based on the logic of argumentation, is described. The model represents argumentation as a directed graph in which nodes and arcs can be colored using an ordinal set of weightings and in which the attributes of both nodes and arcs can be modified. It is thus able to deal with the undercutting or augmenting of arguments. Weightings can be propagated through the graph to generate unique weightings for any node or arc. The model is able to deal with contradiction. It can incorporate numerical methods and is able to handle qualitative and quantitative reasoning.

Representation of Chemical Structures in Knowledge-Based Systems: The StAR System

As part of the StAR project, for the design of a computer system to support risk assessment, it has been necessary to develop a graphical language for the representation of generic structures. These structures are used by rule writers to describe toxicophores for risk assessment. Until now, the input of generic structures in knowledge-based systems has very often been by means of a SMILES-like linear notation. The new StAR graphical language allows the use of a wide range of Markush features including atom lists, bond lists, G-groups, and superatoms. In addition, the StAR graphical language allows rule writers to input rules easily to the knowledge base via a user-friendly graphical interface. This language is well-suited for the representation of chemical features required for expert general toxicity systems, chemical reaction systems, and substructure database systems.

Distinguishing between expert and statistical systems for application under ICH M7

Graphical abstract Figure. No Caption available. HighlightsDefining characteristics of expert and statistical in silico systems are presented.Mechanisms to ensure both systems complement each other are described.Risks of in silico models that may be inappropriate for ICH M7 are highlighted.