Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Comparative data-mining across multiple toxicity databases harmonized by eTOX ontology to test hypotheses

: Steger-Hartmann, T.; Pognan, F.; Sanz, F.; Briggs, K.; Escher, Sylvia; Kleinoeder, T.; Schwab, C.; Wichard, J.; Yang, C.

The Toxicologist 54 (2015), No.1, pp.478, Abstract PS 2226
ISSN: 0731-9193
Society of Toxicology (Annual Meeting) <54, 2015, San Diego/Calif.>
Fraunhofer ITEM ()

The eTOX database has been constructed from diverse preclinical toxicity reports from 13 pharmaceutical companies and continues to be expanded as it exchanges data with other domains. For example, over 100 marketed pharmaceuticals (>10% of the chemicals in the entire eTOX database) have been tested in ToxCast Phase II. The chemical space and associated toxicological findings have been also extended to include cosmetics, food additives, and industrial chemicals by importing databases from COSMOS, HESS, and RepDose into eTOXsys (the user interface of eTOX) database. Based on the greater chemical and biological domains, two use cases were performed to identify and refine chemical categories. The first is based on the hypothesis that steatotic chemotypes (e.g., lipid vacuolization) are similar even for compounds from quite different substance-use spaces (drugs, cosmetics, industrial chemicals). Such chemotypes enable building mechanistic categories for a particular toxicological finding. The second study focused on effects of C6-carboxylic acids and their analogs on specific hematology parameters. The database analysis was able to refute the hypothesis that specific chemical features tend to be associated with a decrease in erythrocytes count and hemoglobin content, increase in reticulocytes, thrombocytes, leucocytes, monocytes and neutrophils counts. It revealed a potential relationship between hematological effects and the pharmacological mode of action. Both data mining processes rely on well-controlled vocabulary of sites and effects. Results from both cases demonstrated that enriching the analysis with information from other substance domains, was critical to explore underlying mode-of-action pathways.