You are here:
Application of IATA - A case study in evaluating the global and local performance of a Bayesian Network model for Skin Sensitization
Citation:
Fitzpatrick, J. AND G. Patlewicz. Application of IATA - A case study in evaluating the global and local performance of a Bayesian Network model for Skin Sensitization. SAR AND QSAR IN ENVIRONMENTAL RESEARCH. Taylor & Francis, Inc., Philadelphia, PA, 28(4):297-310, (2017).
Impact/Purpose:
• Agency Problem: When evaluating potential chemical hazard, the Agency must consider the potential for allergic responses upon skin contact. Skin sensitization is the leading cause of occupational illness in many countries. Skin sensitization data can be generated from in silico, in chemico, in vitro or in vivo approaches. Integration of this information and interpretation for decision making are known as Integrated Approaches to Testing and Assessment (IATA). An IATA for skin sensitization that has been published (Jaworska et al, 2013) • Approach: The current work evaluated the published IATA for skin sensitization. We also characterized the impact of refinements to the IATA. • Results: In the absence of chemical structure indications of skin sensitization, the IATA yielded a skin sensitization potential prediction of 79%. Using chemical structural data, IATA predictivity increased to 89% and our refinements slightly reduced predictivity to 84%less than found for the original model (89%). We found that the original IATA was successful at predicting which chemicals would be skin sensitizers, but not at predicting their relative potency. • Impact to the Agency: The work directly impacts the Agency’s efforts to identify skin sensitizing compounds. This work enables more accurate screening of chemicals.
Description:
Since the publication of the Adverse Outcome Pathway (AOP) for skin sensitization, there have been many efforts to develop systematic approaches to integrate the information generated from different key events for decision making. The types of information characterizing key events in an AOP can be generated from in silico, in chemico, in vitro or in vivo approaches. Integration of this information and interpretation for decision making are known as integrated approaches to testing and assessment or IATA. One such IATA that has been developed was published by Jaworska et al (2013) which describes a Bayesian network model known as ITS-2. The current work evaluated the performance of ITS-2 using a stratified cross validation approach. We also characterized the impact of refinements to the network by replacing the most significant component, the output from a commercial expert system TIMES-SS with structural alert information readily generated from the freely available OECD QSAR Toolbox. Lack of any structural alert flags or TIMES-SS predictions, yielded a sensitization potential prediction of 79% +3%/-4%. If the TIMES-SS prediction was replaced by an indicator for the presence of a structural alert, the network predictivity increased to 84% +2%/-4%, which was only slightly less than found for the original network (89% ±2%). The local applicability domain of the original ITS-2 network was also evaluated using reaction mechanistic domains to better understand what types of chemicals ITS-2 was able to make the best predictions for – i.e. a local validity domain analysis. We ultimately found that the original network was successful at predicting which chemicals would be sensitizers, but not at predicting their relative potency.
