You are here:
THE FUTURE OF COMPUTER-BASED TOXICITY PREDICTION: MECHANISM-BASED MODELS VS. INFORMATION MINING APPROACHES
Citation:
Richard, A M. THE FUTURE OF COMPUTER-BASED TOXICITY PREDICTION: MECHANISM-BASED MODELS VS. INFORMATION MINING APPROACHES. Presented at Nat'l Capital Area Chapter Society of Toxicology -Fall Symp: Computational Approaches for Predicting the Toxicity of Chemicals, Bethesda, MD, 12/11/2000.
Description:
The Future of Computer-Based Toxicity Prediction:
Mechanism-Based Models vs. Information Mining Approaches
When we speak of computer-based toxicity prediction, we are generally referring to a broad array of approaches which rely primarily upon chemical structure information for detemlining a compound's potential for toxicity. The toxicity dimension to this problem includes not only many categories of toxicity of potential interest ( e.g. cancer, immunotox, development/reproductive tox, neurotox, etc ), but within these categories many possible measures for quantitating and representing toxicity (e.g. biochemical, genetic, cellular, in vivo), and multiple components of a toxicity expression in a whole animal (e.g. route of exposure, adsorption, tissue distribution, metabolism, etc.). The prediction and computer modeling dimension to the problem is highly dependent upon the measure of toxicity being modeled, the extent of data and knowledge available relative to the measure, and the biochemical and mechanistic resolution of the toxicity measure. Mechanism-based approaches use chemically and biologically plausible mechanisms to guide model development within well defined classes or regions of chemical-activity space using computational approaches. Information mining approaches, in contrast, take a larger exploratory viewpoint in attempting to process all available data relative to a biological endpoint and extract rules or generalizations based on chemical structure that are sufficient to span the desired prediction space. As different as these two approaches appear, the success of either in meeting the larger goal of toxicity prediction depends on their effective linkage: information mining both generates useful hypotheses for mechanism-based exploration and is a tool for expanding the reach of a mechanism- based model, whereas mechanism-based approaches provide the means for rationalizing empirical correlations or generalizations based on the underlying reaction chemistry and biochemical mechanisms, providing the scientific foundation for further model development and extrapolation. A series of examples will be provided to illustrate how these two approaches to the investigation of toxicity can complement each other and enhance our ability to understand and predict chemical toxicity.
This abstract does not necessarily reflect EP A policy.