You are here:
Predictive Modeling of Apical Toxicity Endpoints Using Data From ToxCast
JUDSON, R., D. J. DIX, K. A. HOUCK, M. MARTIN, R. J. KAVLOCK, I. A. SHAH, AND T. B. KNUDSEN. Predictive Modeling of Apical Toxicity Endpoints Using Data From ToxCast. Presented at Society of Risk Assessment Annual Meeting, Boston, MA, December 07 - 10, 2008.
The main goal of ToxCast is the discovery and validation of “signatures” linking in vitro assay data to in vivo toxicity endpoints. These signatures will be collections of assays that are correlated with particular endpoints. These assay collections should also help define molecular-and cellular-level mechanisms of toxicity. This talk will discuss our strategy to use a combination of statistical and machine learning methods, coupled with biochemical network or systems biology approaches. Our initial examples will focus signatures for endpoints from 2 year rodent cancer bioassays. Most of the data we have analyzed is in dose or concentration response series, so to effectively use this data we have developed novel approaches to combine many kinds of dose-response data together with standard machine learning methods. A key issue to be discussed is the validation of ToxCast predictive signatures, an issue involving statistics, as well as data coverage in both biological and chemical space.
The US EPA and other regulatory agencies face a daunting challenge of evaluating potential toxicity for tens of thousands of environmental chemicals about which little is currently known. The EPA’s ToxCast program is testing a novel approach to this problem by screening compounds using a variety of in vitro assays and using the results to prioritize chemicals for further, more detailed testing. Phase I of ToxCast is testing 320 chemicals (mainly pesticide active ingredients) against ~400 cell-based and biochemical assays. In order to anchor these studies, we are using in vivo guideline study data for subchronic, chronic, cancer, reproductive and developmental endpoints. This data is compiled in the EPA toxicity reference database, ToxRefDB.