You are here:
Evaluation of Existing QSAR Models and Structural Alerts and Development of New Consensus Models for Genotoxicity Using a Newly Compiled Experimental Dataset
Citation:
Pradeep, P., R. Judson, AND G. Patlewicz. Evaluation of Existing QSAR Models and Structural Alerts and Development of New Consensus Models for Genotoxicity Using a Newly Compiled Experimental Dataset. ASCCT Annual meeting, Virtual, NC, October 20 - 22, 2020.
Impact/Purpose:
Abstract submitted to the ASCCT 2020 meeting in Oct 2020. Integrated testing and assessment approaches (IATA) are applied to evaluate the genotoxic potential of chemicals using a combination of in silico, in vitro and in vivo approaches. These models will provide a robust support framework for assessing genotoxicity potential for new and untested chemicals.
Description:
Genotoxicity is among the toxicological endpoints that pose the highest concern for human health and is subject to regulatory assessment. Integrated testing and assessment approaches (IATA) are applied to evaluate the genotoxic potential of chemicals using a combination of in silico, in vitro and in vivo approaches. A major effort was undertaken to compile data from several sources (TOXNET, COSMOS, eChemPortal and ECVAM) and harmonize the names and outcomes of different assay types (bacterial mutagenicity (Ames), chromosomal aberrations (Clastogen), and others). The data was evaluated using a conservative IATA to assess genotoxic potential using the classification scheme by Williams et al., 2019. The dataset comprised 4828 chemicals, of which 2553 chemicals were categorized as genotoxic and 1819 as non-genotoxic using the IATA. The IATA assigned a chemical as genotoxic if any single Ames or Clastogen assay was positive; ~16% chemicals active in <50% Ames assays were classified as genotoxic and ~15% chemicals with inconclusive Ames data were classified as genotoxic. So, a new cut-off-based classification scheme was developed where genotoxicity classifications were made based on percentage of Ames and Clastogen assays a chemical was active in. Next, QSAR tools (TEST and Lazar) and the OECD Toolbox structural alerts/profilers (e.g. OASIS DNA alerts for Ames, CA) were used to make in silico predictions for genotoxicity. The performance of individual QSAR tools and alerts was evaluated against the IATA and newly defined cut-off-based genotoxicity classifications. The balanced accuracies ranged from 64-80%. Finally, a naïve Bayes consensus model was developed using two combinations of QSAR tools and alert predictions. The consensus models do not result in significant differences in the overall prediction across various combinations, with balanced accuracies ranging from 50-76%. Overall, the predictivity of individual tools and the consensus models is slightly improved using the experimental activity cut-off-based classification scheme relative to the Williams et al., 2019 scheme for genotoxicity prediction. These models will provide a robust support framework for assessing genotoxicity potential for new and untested chemicals. This abstract does not necessarily represent U.S. EPA policy.