You are here:
Novel Informatic Approaches to Analyze Gene Expression Data with the ToxCast 320 Chemical Library in Cultures of Primary Human Hepatocytes
Citation:
Beam, A., D. Rotroff, K. Freeman, A. Farmer, H. Bondell, K. A. HOUCK, R. JUDSON, D. J. DIX, E. LeCluyse, AND S. Ferguson. Novel Informatic Approaches to Analyze Gene Expression Data with the ToxCast 320 Chemical Library in Cultures of Primary Human Hepatocytes. Presented at Society of Toxicology Annual Meeting, Baltimore, MD, March 15 - 19, 2009.
Impact/Purpose:
Unique to this approach is assessment of concentration-response changes over time (6, 24, 48 hr in culture) as well as correlation of gene targets with one another. From these analyses, inclusion of data from all time points resulted in more accurate clustering of the replicate ToxCast 320 and reference chemicals that reduced donor-dependent variability. This approach has significant implications in standardizing primary hepatocyte data analysis across donors and profiling chemical response with in vivo endpoints.
Description:
Prevailing methodologies in the analysis of gene expression data often neglect to incorporate full concentration and time response due to limitations in throughput and sensitivity with traditional microarray approaches. We have developed a high throughput assay suite using primary human hepatocyte cultures as in vitro models that retain liver-like functionality (e.g. induction, metabolism and transport) to generate more comprehensive data across time and concentration. Using the ToxCast 320 chemical library, mRNA expression was determined using a quantitative nuclease protection assay using the Omix Imaging System (HTG, Tucson, AZ). Fourteen gene targets representing Phase I/II metabolism and transport were monitored based on their role in liver function and sensitivity to receptor pathways (AhR, CAR, PXR, PPARα, FXR). Techniques from machine learning were used to cluster compounds by gene response profiles. Dose-responses were mathematically abstracted as vectors in multidimensional space (rather than classical scalar representations traditionally associated with standard microarray analyses) and used in algorithms such as K-means and algometric clustering to create representative chemical phylogenies.