You are here:
Uncertainty in QSAR models with variability in experimental data
Citation:
Pradeep, P. Uncertainty in QSAR models with variability in experimental data. Presented at Society of Toxicology Annual Meeting, Baltimore, Maryland, March 10 - 14, 2019.
Impact/Purpose:
Session Proposal and abstract for SOT 2019 annual meeting
Description:
New approach methodologies (NAMs), such as quantitative structure activity relationship (QSAR) models based on chemical structure information, are commonly used to predict hazard in the absence of experimental data. QSAR models are developed and validated using experimental (i.e., in vitro or in vivo) toxicity data. However, variability in the experimental data leads to uncertainty in QSAR model predictions and impacts model quality estimates. To accurately characterize model quality, QSAR models should be benchmarked according to the variability and uncertainty in the experimental studies. This talk will present examples demonstrating limitations on model predictivity owing to data variability and present methods to quantify model uncertainty driven by data variability using two sets of QSAR models. The first set of models integrate the QSAR modeling technique with the toxic equivalency factors (TEFs) scheme to predict neurotoxic equivalency factors for polychlorinated biphenyls (PCBs) using scaffold-based chlorine substitutions as a chemical fingerprint. The correlation analysis of underlying data sources and the model performance demonstrates that large uncertainties in the underlying data result in QSAR models that are limited in their predictivity. The second set of models were developed to predict in vivo points of departure (POD, the point on the dose-response that marks the beginning of a low-dose extrapolation) using chemical structural and physicochemical properties. These models demonstrate the use of a bootstrapping technique to incorporate typical lab to lab variability (~0.5 log-units) in the POD values for deriving confidence intervals in model predictions. These examples illustrate the effect of variability in experimental data on uncertainty in QSAR model predictions. This abstract does not necessarily reflect U.S. EPA policy.