Citation:

Schmieder, P., Rick Kolanczyk, M. Hornung, M. Tapper, J. Denny, B. Sheedy, AND H. Aladjov. A rule-based expert system for chemical prioritization using effects-based chemical categories. SAR AND QSAR IN ENVIRONMENTAL RESEARCH. Taylor & Francis, Inc., Philadelphia, PA, 25(4):253-287, (2014).

Impact/Purpose:

The ER expert system was built for chemical prioritization of fooduse inert ingredients and antimicrobial pesticides. The work has been done in conjunction with OCSPP and has been extensively reviewed by two EPA FIFRA SAPs in 2009 and more recently in January 2013.

Description:

A rule-based expert system (ES) was developed to predict chemical binding to the estrogen receptor (ER) patterned on the research approaches championed by Gilman Veith to whom this article and journal issue are dedicated. The ERES was built to be mechanistically-transparent and meet the needs of a specific application i.e., predict for all chemicals within two well-defined inventories (industrial chemicals used as pesticide inerts, and antimicrobial pesticides) all lacking structural features associated with high affinity binders. Similar to the high-quality fathead minnow database upon which Veith QSARs were built, the ERES was derived from what has been termed gold standard data, systematically collected in assays optimized to detect even low affinity binding of industrial chemicals thus maximizing confidence in the negatives. The resultant logic-based decision tree ES, determined to be a robust model, contains seven major nodes with multiple effects-based chemicals categories within each. Rules are formulated from test data combined into categories within interaction mechanisms. Predicted results are presented in the context of empirical data within local chemical structural groups facilitating informed decision making. Even with optimized detection, the ERES applied to two inventories of > 600 chemicals resulted in only ~5% of the chemicals predicted to bind ER.

URLs/Downloads:

Record Details: