You are here:
20190925 - Evaluation of the Predictive Accuracy of QSAR Models and Alerts for Genotoxicity Using a Newly Compiled Experimental Dataset (ASCCT)
Citation:
Pradeep, P., R. Judson, AND G. Patlewicz. 20190925 - Evaluation of the Predictive Accuracy of QSAR Models and Alerts for Genotoxicity Using a Newly Compiled Experimental Dataset (ASCCT). American Society for Cellular and Computational Toxicology (ASCCT) annual meeting, Gaithersburg, MD, September 25 - 26, 2019. https://doi.org/10.23645/epacomptox.9924524
Impact/Purpose:
Poster presented to the American Society for Cellular and Computational Toxicology (ASCCT) annual meeting in September 2019. The 8th annual conference theme is 'Computational Toxicology: Peeking into the Clouds while Keeping our Feet on Solid Ground'.
Description:
Carcinogenicity and mutagenicity are among the toxicological end points that pose the highest concern for human health and are subject to regulatory assessment. Integrated approaches to testing and assessment (IATA) are applied to evaluate the genotoxic potential of chemicals using a combination of in silico, in vitro and in vivo approaches. Here, the predictive accuracy of several publicly available genotoxicity QSARs and structural alerts were assessed using a large new dataset. Data from assays that detect bacterial mutagenicity (Ames) or chromosomal aberrations (CA) were evaluated using a conservative IATA to derive a call for genotoxic potential (Cross et al., 2019). The IATA assigned a chemical as genotoxic, if any single assay was positive. However, for ~30% of chemicals, this conservative call disagreed with the consensus call (e.g. a chemical with <50% Ames positive assays was still assigned ‘positive’). This highlights the need for further data curation. Using the IATA, the dataset comprised 4828 chemicals, of which 2553 chemicals were categorized as genotoxic, 1819 as non-genotoxic. Toxicity Estimation Software Tool (TEST) and Lazar (Lazy structure–activity relationships) were used to predict Ames, the OECD Toolbox to identify presence of structural alerts using 5 profilers (e.g. OASIS DNA alerts for Ames, CA). The in silico predictions were compared against the experimental genotoxicity calls. Balanced accuracies varied from 64-78%, with some QSARs/alerts having low sensitivity but high specificity and vice versa. Next steps include further data curation and development of new consensus models of genotoxicity. This abstract does not necessarily represent U.S. EPA policy.
URLs/Downloads:
DOI: 20190925 - Evaluation of the Predictive Accuracy of QSAR Models and Alerts for Genotoxicity Using a Newly Compiled Experimental Dataset (ASCCT)
ASCCT-PPRADEEP-V4.PDF (PDF, NA pp, 783.122 KB, about PDF)