Home > Project > Background and Concept Print page Send page

General Information   Summary & Objectives   Background & Concept   Potential Impact of Project Results   Project Activities

 

 

 

 

 

Background and Concept

 

The reliable prediction of toxic effects of human relevance with the help of computer systems, be it with QSAR or rule-based systems, is currently mainly restricted to the endpoints mutagenicity and carcinogenicity and, to a lesser extent, (skin) sensitisation. The reason for this focus is twofold: first the mechanisms leading to these toxic effects are relatively well-characterized and of a limited complexity (in the case of carcinogenesis this is mainly true for genotoxic carcinogens); secondly, there is abundant literature in the public domain for such endpoints, which can be used to construct expert rule or quantitative-structure activity relationships.

 

There have been several attempts to model and predict more complex toxic endpoints on the basis of available in vivo toxicity data (e.g. MTAs or CRADAs within the US Food and Drug Administration [FDA] on the one hand, and different software providers such as MDL Information System / MDL QSAR, Leadscope, Multicase, Bioreason etc. on the other hand, for the prediction of teratogenicity, reproductive toxicity, 90-day organ toxicity or maximum recommended therapeutic dose), or on the basis of gene expression data (e.g. Biowisdom, ToxWiz). However, no major breakthrough has been discernable up to now, i.e. there is no broad acceptance and usage of these systems or models.

 

The main reasons for this scenario are:

 

 

The endpoint to be modelled is far too complex in order to be displayed in one global model or a single structure-activity relationship (e.g. hepatotoxicity). The available models are constructed to identify hazard for a particular chemical (e.g. liver or kidney toxicity) but are worthless for risk assessment in the context of drug development without dose-effect relationship. There is a lack of sufficient sets of toxicological data showing reliable quality and quantity, and including structural information, to build robust models.

 

 

The proposed eTOX project targets to overcome these two main hurdles for in vivo toxicity prediction by:

 

 

sharing of high quality proprietary toxicity data. The majority of the shared data will originate from GLP or GLP-like studies and will represent the fundament for all future prediction efforts. These data will come from the archives of EFPIA companies and were previously accessible for modelling or rule base development only for the owning company. A database with harmonized ontologies will be created to store these data. development of several models representing the different components in the mechanism leading to a toxic effect (including DMPK), which will then be integrated in an overall decision-making tool allowing prediction of in vivo toxicity

 

 

A major breakthrough may be expected through this approach, leading to predictions of in vivo toxicity with reasonable reliability. Even if the development of a new system to predict in vivo toxicity results less successful than expected, the project will result in a high quality toxicity database available to all participants for future reference and data analysis.

A potential breakthrough, i.e. the intended advantages of the eTOX prediction system, will be demonstrated by the verification and validation of the new system in comparison with existing commercial systems. The ultimate measure for determining the success of the eTOX system would be a lowering in the attrition rates during the drug development process or a lowered rate of adverse drug reactions in clinical trials. However, this measure will most probably be available only several years after the completion of the project