S2: Interpreting Toxicogenomics Data

Analyzing and interpreting toxicogenomics data at the network level
Ralf Herwig, Max-Planck-Institute for Molecular Genetics, Berlin

Ralf Herwig


Max-Planck-Institute for Molecular Genetics, Berlin


Department of Computational Molecular Biology


A major goal of human in vitro toxicogenomics is to improve the assessment of the toxic hazard (and ultimately the risk) of drugs, chemicals and other compounds with assays based on human cell lines rather than in vivo in rodents. In particular, gene expression read-outs are used in order to identify responsive genes and biological pathways capable of predicting toxicity1-4.

Here, we present results achieved for human in vitro toxicogenomics within several projects funded by the European Commission (carcinoGENOMICS, diXA, HeCaToS).

Among others, we present a computational method that maps expression data onto molecular pathways and uses these pathways to discriminate different toxicity classes. Using this approach, we showed, for example that in vitro hepatocyte-like assays derived from human embryonic stem cells can be utilized for carcinogenic hazard assessment of chemicals and that the approach based on pathway response patterns performs better than the classical gene pattern approach2.

This result holds as well for other in vitro assays3.

Since the approach builds on pre-annotated human pathways, we introduce a computational resource for human molecular interactions, the ConsensusPathDB, that integrates the content of thirty publicly available interaction databases and that holds functionality for interpreting high-throughput data at the network level5-6.