You are here:
COMPUTATIONAL TOXICOLOGY ADVANCES: EMERGING CAPABILITIES FOR DATA EXPLORATION AND SAR MODEL DEVELOPMENT
Citation:
Richard, A M. AND C. R. Williams. COMPUTATIONAL TOXICOLOGY ADVANCES: EMERGING CAPABILITIES FOR DATA EXPLORATION AND SAR MODEL DEVELOPMENT. Presented at Environmental Mutagen Society Meeting, Symposium on "Innovation in Toxicological Sciences", Miami Beach, FL, May 10-14-2003.
Description:
Computational Toxicology Advances: Emerging capabilities for data exploration and SAR model development
Ann M. Richard and ClarLynda R. Williams, National Health & Environmental Effects Research Laboratory, US EPA, Research Triangle Park, NC, USA; email: richard.ann@epa.gov
As a key capability in the computational toxicology arsenal, structure-activity relationship (SAR) methods play an essential role in strategies to screen and prioritize chemicals for a wide range of potential toxicities, including carcinogenicity and mutagenicity. However, current SAR capabilities are circumscribed by the nature, scope, and restricted availability of existing toxicity data. Trends in this field that have the potential to advance SAR to a new level of utility include: 1) new data,i.e. migrating more data into the public domain and expanding SAR training sets to include chemicals spanning a larger range of structure-activity space (e.g., pharmaceuticals); 2) reformulations of old data,i.e. refining definitions of biological activities from summarized data and protocols to better achieve the prediction objectives of the SAR study; 3) new approaches, i.e. data mining algorithms that provide for dynamic exploration of global datasets to create customized class-based models on-the-fly; 4) improved public access and structural annotation for existing data,i.e. improving structure-linkages and standards for existing toxicity databases, and converting databases to these standards to enable unrestricted access and exploration of existing data from an SAR perspective; and 5) data integration,i.e. with the advances in 4) the opportunity for much broader integration of historical toxicity data with 'omics data becomes possible, moving SAR into the realm of chemo-bioinformatics. Examples of advances in each of these areas will be given. In particular, progress will be reported on the development of the DSSTox (Distributed Structure-Searchable Toxicity) database network for historical toxicity data and a companion TOXML initiative to create standard schema for new toxicity data collection. The joint goals of these projects are to promote toxicity data file standardization, controlled vocabularies, full data file access, and inclusion of chemical structures and structure-searching capabilities within and across public toxicity databases. This abstract does not reflect EPA policy.