Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

New strategies for the generation of chemical categories under REACH

: Batke, Monika; Gundert-Remy, U.; Helma, C.; Kramer, S.; Kleppe-Nordqvist, Sara; Maunz, A.; Partosch, F.; Seeland, M.; Bitsch, Annette

Naunyn-Schmiedebergs archives of pharmacology 386 (2013), Supplement 1, pp.S6
ISSN: 0028-1298
ISSN: 1432-1912
Deutsche Gesellschaft für Experimentelle und Klinische Pharmakologie und Toxikologie (Annual Meeting) <79, 2013, Halle/Saale>
Journal Article, Conference Paper
Fraunhofer ITEM ()

Although within the REACH regulation a comprehensive toxicological data set is required for risk assessment it is stipulated that animal experiments are to be avoided whenever possible. Thus, methods that allow predicting the inherent toxic properties of chemical substances are needed. The aim of the present project is to develop an innovative strategy for setting up categories that will enable a toxicological assessment for repeated-dose administration. Based on published in vivo studies, identical toxicological properties of chemicals (toxicological fingerprinting) in combination with similarity in the chemical structure shall be used to group substances in categories. This 2-dimensional matrix will then allow the toxic properties of untested chemical materials to be estimated. The prediction rules will be implemented in a publicly available product that is able to assign chemicals to categories and thus maps to anticipated toxicity insilico.
The data basis for these developments consists of two independent data bases for industrial chemicals (RepDose and Neustoff). In a first step those data which are suitable had to be extracted in a so called homogeneous subset and for any exclusion the criteria had to be clearly defined. In a second step it was decided to merge both data bases at this point to enhance the applicability domain. Two independent sets were created in a random selection process - one for model development and a second one for validation. At present a dataset of about 800 chemicals tested in more than 900 studies is available for modeling. On this basis data clustering is performed on structural and on toxicological basis at the moment. Recurring problems and discussion points in the clustering process is the handling of "missing values". Different clustering strategies are followed so far and software is available which can be shared by all project partners.
Furthermore, prediction models for specific organ toxicity are evaluated. Refinement of the datasets and continuous adaptation of the processes is ongoing work at the moment.