You are here:
Mining a human transcriptome database for chemical modulators of Nrf2
Citation:
Rooney, J., S. Hiemstra, B. Chorley, B. van de Water, AND C. Corton. Mining a human transcriptome database for chemical modulators of Nrf2. World Congress on alternatives and animal use in the life sciences, Seattle, Washington, August 20 - 24, 2017.
Impact/Purpose:
EPA is responsible for regulating tens of thousands of chemicals, most of which have little or no toxicology data. EPA cannot test all of these chemicals in animal models due to cost and time. Faster, less resource rich methods are needed to evaluate chemicals especially using methods that can put the pathways affected into the context of adverse outcome pathways. In the present study, computational procedures were developed to enable the identification of chemical modulators of Nrf2 in a large database of human microarray data (~38K profiles).
Description:
Computational procedures were developed to enable the identification of chemical modulators of Nrf2 in a large database of human microarray data (~38K profiles). Ten gene expression biomarkers were constructed from microarray experiments in which human hepatocyte, breast, and lung tissues were exposed to known Nrf2 activators or in which the Nrf2 suppressor Keap1 was knocked down. Biomarker genes were identified that exhibited consistent directional expression changes and were altered in the opposite direction upon Nrf2 knockdown. As validation, these preliminary biomarkers were tested for predictive accuracy by comparing to ~60 microarray comparisons (biosets) representing 35 chemicals with known Nrf2 activity. Comparisons were carried out using the Running Fisher algorithm, a statistical test of pair-wise similarity in expression. The most predictive biomarker (93% predictive accuracy) was used to assess Nrf2 activity across 10,171 biosets examining the effects of 1963 chemicals of unknown Nrf2 activity. A total of 249 and 44 chemicals were found to activate or suppress Nrf2, respectively. Predictions are now being independently validated using a Nrf2 reporter. Using Nrf2 as a proof-of-concept, we provide a computational method for mining large human transcriptome databases to find chemical modulators of transcription factors. These in silico methods can be used to populate adverse outcome pathway networks by mechanistically linking key events and adverse outcomes. (This abstract does not reflect EPA policy.)