You are here:
Mode of action from dose-response microarray data: case study using 10 environmental chemicals
Citation:
PIRONE, J. R., D. J. DIX, M. J. DEVITO, AND I. A. SHAH. Mode of action from dose-response microarray data: case study using 10 environmental chemicals. Presented at Society of Toxicology Annual Meeting, Baltimore, MD, March 15 - 19, 2009.
Impact/Purpose:
A set of curve shape and intensity features was extracted from the fit for each gene found to exhibit a significant dose-response. Features included first-derivatives, maximum/minimum expression, dose at maximum/minimum expression, and area under the curve. Using subsets of curve features, genes were first partitioned into groups that responded with a similar shape then into groups with similar expression intensities. Partitioning was performed on a chemical by chemical basis. Gene groups were subjected to functional and pathway analysis.
Description:
Ligand-activated nuclear receptors regulate many biological processes through complex interactions with biological macromolecules. Certain xenobiotics alter nuclear receptor signaling through direct or indirect interactions. Defining the mode of action of such xenobiotics is difficult due to the many perturbations in the cellular signaling networks resulting from exposure. Microarray data, when collected in a dose-response setting, is a rich source of information for determining the MOA. Analysis presents several challenges: namely, it is difficult to choose a single quantitative model that can be applied to each gene. We have developed a method for analyzing microarray dose-response data that is flexible, while still capable of capturing the complexity of the responses. Rat primary hepatocytes were incubated with solutions of 0, 10, 30, and 100 μM of 4-nonylphenol, MEHP, myclobutanil, propiconazole, or triadimefon or 1, 3, and 1 μM of DE-71, PCB-118, or PCB-153. In all cases, the control consisted of the 1% DMSO dosing solution. RNA was extracted 72 hrs after dosing and analyzed using Affymetrix Rat Genome 230 2.0 arrays. Natural cubic splines were fit to the dose-response data for each gene. Statistical significance of each fit was assessed using bootstrap analysis.