Strategies and challenges in predictive toxicology


Glenn J. Myatt, Paul E. Blower, Kevin P. Cross, Wayne P. Johnson, and Chihae Yang. Leadscope, Inc, 1393 Dublin Road, Columbus, OH 43215
There are numerous challenges to the development of a comprehensive strategy for predictive toxicology. Access to high quality data from which accurate predictive models can be generated, continues to be a major impediment. An approach to domain-intelligent integration of disparate sources, both electronic and non-electronic, will be described. A toxicology controlled vocabulary, ToxML, based on the XML standard is central to the integration. Prior to building any predictive model, an assessment of the data is required to transform complex hierarchical XML data into decision point data. This process usually involves an expert judgment. Only once the data has been selected, integrated and assessed can model building start and even then there is no guarantee that the data can be modeled. An approach to building and applying predictive model will be described from various stages of the workflow, the data integration, assessment, and subsetting to prepare modelable data set.

Herman Skolnik Award Symposium
8:15 AM-11:45 AM, Tuesday, August 24, 2004 Pennsylvania Convention Center -- 110A&B, Oral

Division of Chemical Information

The 228th ACS National Meeting, in Philadelphia, PA, August 22-26, 2004