The use of omics-based approaches in regulatory toxicology: an alternative approach to assess the no observed transcriptional effect level

The evaluation of chemical hazard is based on the identification of the quality and the quantity of adverse effects as a consequence of exposure. The adverse effects that do not involve genetic damage are often related to the chemical dose or concentration. The adverse outcome is the consequence of a row of key events, each targeting a different biological trait. The identification of these key events at molecular and cellular level would provide novel biomarkers of exposure and risk. The application of toxicogenomics approaches to experimental models of chemical exposure allows the detection of gene pathways involved in response to low doses of the chemical as an early endpoint of adversity. The use of toxicogenomics would improve the knowledge on the dose-response relationship, linking the environmental exposure to the effect on the population and allowing a better refinement of the quantitative risk assessment. In this context, the gene modulation data can be used to calculate a No Observed Transcriptional Effect Level (NOTEL). In this paper we present a method for evaluating the NOTEL based on anomaly detection: a classifier is built that discriminates between target class instances, i.e., normal cases, and anomalies, i.e., samples with significant transcriptional effects. The strength of this method is that (i) it can be applied to cases in which few samples are available and the space dimension is high and (ii) it makes use of the complete gene expression profiles.