You are here:
Prioritizing ToxCast Chemicals Across Multiple Sectors of Toxicity Using ToxPi
Citation:
REIF, D., M. T. MARTIN, H. M. MORTENSEN, S. GANGWAL, A. WANG, R. JUDSON, P. VOLARATH, R. J. KAVLOCK, AND D. J. DIX. Prioritizing ToxCast Chemicals Across Multiple Sectors of Toxicity Using ToxPi. Presented at Annual Meeting of the Society of Toxicology, Washington, DC, March 06 - 10, 2011.
Impact/Purpose:
ToxPi consolidates information from multiple domains—including ToxCast™ in vitro bioactivity profiles (a wide-ranging battery of over 500 high-throughput screening assays), inferred toxicity pathways, exposure predictions, and chemical properties/descriptors—into comprehensive toxicity scores and multivariate visualizations representing the contribution of each data domain to overall priority rankings.
Description:
The Toxicological Prioritization Index (ToxPi™) framework was developed as a decision-support tool to aid in the rational prioritization of chemicals for integrated toxicity testing. ToxPi consolidates information from multiple domains—including ToxCast™ in vitro bioactivity profiles (a wide-ranging battery of over 500 high-throughput screening assays), inferred toxicity pathways, exposure predictions, and chemical properties/descriptors—into comprehensive toxicity scores and multivariate visualizations representing the contribution of each data domain to overall priority rankings. Here, we develop a generalizable methodology for aligning ToxCast assay data with diverse prioritization tasks and demonstrate an implementation for profiling and prioritizing chemicals according to four sectors of toxicological concern: systemic (non-cancer), cancer, developmental, and reproductive. The methodology draws upon the wealth of knowledge in existing knowledgebases to link biological pathways with relevant in vitro assay results. The multi-sector ToxPi profiles developed here can be viewed as an overall prioritization or decomposed into individual sectors for targeted evaluation. Unsupervised clustering of the ToxPi profiles segregated chemicals into sector-specific groups of toxicological concern, which revealed previously unrecognized clusters of chemicals having similar patterns of bioactivity. Exploring these chemical clusters in the context of in vivo data from ToxRefDB™ suggested testable hypotheses about the potential toxicity of chemicals without in vivo data. These results indicate that an overall weight-of-evidence approach that retains transparency as to the contribution of individual data sources can be useful in supporting prioritization decisions across diverse areas of toxicological concern. (This abstract does not necessarily reflect Agency policy.)