January 26, 2016
Many commercial and environmental chemicals lack toxicity data necessary for users and risk assessors to make informed decisions about their potential health effects. Computational methods use data about structure, properties, and toxicity from tested chemicals to make predictions about the characteristics of untested chemicals. These include quantitative structure-activity relationship (QSAR) models, which predict the activities of chemicals with unknown properties by relating them to properties of known chemicals, and read-across, which uses toxicity data from a known (source) chemical to predict toxicity for another (target) chemical, usually but not always on the basis of structural similarity. Predictions made using these methods about toxicity of untested chemicals can help set priorities for future in vitro or in vivo testing, ensuring that the most important hazards are characterized first and that testing resources are used efficiently.
Fundamentals of QSAR Modeling: Basic Concepts and Applications
Alex Tropsha, Ph.D., University of North Carolina at Chapel Hill
Tropsha explains how QSAR models allow chemical compounds to be characterized mathematically, enabling statistical predictions about the properties of untested chemicals. He emphasizes the importance of curation of chemical and biological databases, points out some common errors in QSAR development, and reviews case studies in which QSARs were used to predict toxicities such as skin sensitization and liver toxicity.
Application of QSAR Principles in the Regulatory Environment: The U.S. EPA New Chemicals Program
Louis (Gino) Scarano, Ph.D., Office of Pollution Prevention and Toxics, U.S. Environmental Protection Agency
Scarano reviews how the EPA uses quantitative tools and models to generate predictions about carcinogenicity, exposure potential, aquatic toxicity, and other toxic effects of new chemicals. These predictions are then used to make occupational risk assessments and identify where further testing might be needed.