The Toxicologist - Society of Toxicology

calculated using this new tool are generally consistent with those calculated by
Rodgers et al., while providing a convenient, approachable and high-throughput
tool to estimate these measures for a large variety of xenobiotics.
1810 A COMPUTATIONAL FRAMEWORK FOR
CHARACTERIZING BIOMARKERS OF
ORGANOPHOSPHORUS INSECTICIDE MIXTURE
EXPOSURE.
J. H. Ivy1, 2 , J. M. Wright2 , J. L. Rogers1, 2 , A. N. Mayeno1, 2 , M. A. Lyons1, 2
and B. Reisfeld1, 2 . 1Department of Chemical and Biological Engineering, Colorado
State University, Fort Collins, CO and 2Systems and Computational Biology Research
Group, Colorado State University, Fort Collins, CO.
The disposition and toxicological effects of organophosphorus insecticides (OPs)
following human exposure are governed by complex absorption, distribution, metabolism,
and elimination (ADME) and pharmacodynamic (PD) processes. Among
the signatures of these processes are a variety of metabolites, some of which are
common to many OPs and some of which are unique to a given OP. Characterizing
these biomarkers of exposure and effect, and their quantitative relationship to the
an individual’s ADME and PD, warrants the development and use of quantitative
and computational modeling tools. Here we describe the development and usage of
such a modeling tool that can facilitate the analysis of tissue dosimetry based on
given dosing scenarios and reconstruct dose from specific biomarker data. This software
framework, DoseSim:OP, is centered around a sophisticated differential equation
solution and statistical simulation engine, allowing analyses that include
Bayesian inference and Monte Carlo stochastic simulations. The modular architecture
behind DoseSim:OP allows different models (ADME+PD) to be linked to the
simulation engine, allowing great flexibility in the systems to be analyzed. The first
of the models to be implemented is a physiologically-based pharmacokinetic model
for a mixture of chlorpyrifos [O,O-diethyl O-(3,5,6-trichloro-2-pyridyl)phosphorothioate]
and diazinon [O,O-diethyl O-(2-isopropyl-6-methyl-4pyrimidinyl)phosphorothioate]
based upon the work of Timchalk et al. (2008). We
have further refined this model using data from our own studies, coupled with optimizations
and global sensitivity analyses. Predictions from DoseSim:OP for various
exposure and dose reconstruction scenarios have shown good agreement with
those from the literature and to our own experiments, and demonstrate the flexibility
and utility of this methodology and framework. This project is supported by
EPA STAR Grant #RE83345101.
1811 QSAR ANALYSIS OF THE ATSDR DATABASE OF
CHEMICAL HEALTH GUIDANCE VALUES.
E. Demchuk 1 , Y. Tie 1 , T. Miller 2 and R. M. Garrett 2 . 1 Division of Toxicology &
Environmental Medicine, ATSDR/CDC, Atlanta, GA and 2 Department of Defense,
Washington, DC. Sponsor: B. Fowler.
The ATSDR SEQUOIA (formerly HazDat) database contains information concerning
thousands of chemicals collected at hazardous waste sites across the nation.
However, less than 10% of them have been extensively reviewed to provide public
health guidance. A goal of this project is to develop QSAR (Quantitative Structure
Activity Relationship) models using these well-studied chemicals and apply the
models to estimate provisional health guidance values (pHGVs) for other toxic
chemicals that have not yet been reviewed by ATSDR. The QSAR analysis was conducted
on a set of 100 compounds with rat oral chronic LOAEL (lowest-observedadverse-effect-level)
values; and a QSAR model with R2 and Q2 of 0.92 and 0.82,
respectively, was developed. The model was applied to assess the toxicity of n-alkanes
found in the 2010 BP Gulf oil spill, because limited toxicological information
is available for long-chain alkanes and their mixtures. The LOAEL estimates obtained
using our model were in fair agreement with those obtained using the TOP-
KAT package, and they were in an excellent agreement with information on
ADMET properties of n-alkanes found in the ATSDR toxicological profile for
Total Petroleum Hydrocarbons. Further validation of the model and expansion of
its chemical space domain are under way. In the future, a robust model for pHGVs
may complement the risk assessment of pure chemicals and multi-component exposures
carried out using conventional methods.
1812 SUPPORTING SAFETY ASSESSMENT OF DRUG
IMPURITIES THROUGH EXAMINATION OF AN AMES
ASSAY QSAR MODEL.
G. Myatt 1 , K. P. Cross 1 and L. G. Valerio 2 . 1 Leadscope Inc., Columbus, OH and
2 CDER/OPS, U.S. FDA, Silver Spring, MD. Sponsor: N. Sadrieh.
Drugs under development may contain multiple impurities with structural alerts
portending to toxicity, and thus to public health. According to the FDA draft guidance
to industry on genotoxic and carcinogenic impurities in drug products and
388 SOT 2011 ANNUAL MEETING
substances, impurities can be evaluated for genotoxicity based on structure-activity
relationship (SAR) analysis. These molecules if predicted positive may need to have
their levels controlled or the impurity synthesized and qualified in the Ames assay.
Given the importance of adequately qualifying genotoxic impurities, this study
evaluated a quantitative SAR (QSAR) analysis model developed that predicts Ames
assay outcomes based on a training dataset of over 3000 molecules. A drug-like external
validation set of over 3000 molecules was used to characterize the QSAR
model performance by structural class, therapeutic category, and domain-of-applicability.
Preliminary predictive performance statistics for the external test set (79%
sensitivity, 76% specificity, 77% accuracy) were comparable to the model cross-validation
statistics (77% sensitivity, 88% specificity, 83% accuracy). Hundreds of
drug impurity molecules (proprietary and public) were examined at the FDA to assess
the suitability of the chemical space of the Ames model for predicting drug
molecules. The predominant molecular features contributing to Ames positive predictions
of the drug impurities were elucidated through hierarchical classification of
discriminating model features. This analysis has led to an improvement for predicting
the mutagenicity of drug impurities in the Salmonella assay and by extending
our understanding of known and new structural alerts in the Ames assay. The
QSAR model is intended to help support decision-making during safety evaluations
of the genotoxic potential of drug impurities.
1813 ESTIMATING TOXICITY-RELATED BIOLOGICAL
PATHWAY ALTERING DOSES FOR HIGH-
THROUGHPUT CHEMICAL RISK ASSESSMENTS.
R. Judson, D. J. Dix, K. J. Robert, R. Setzer, E. A. Cohen Hubal and M. T.
Martin. U.S. EPA, Research Triangle Park, NC.
The doses at which toxicity-related biological pathways are altered, measured using
in vitro assays, are analogs to regulatory values such as Reference Doses (RfD) derived
from in vivo toxicity tests. These biological pathway altering doses (BPADs)
could be used in high throughput risk assessments of 1000s of data-poor environmental
chemicals. Estimating a BPAD requires definition of biological pathways
that can lead to toxicity, and development of in vitro assays for these pathways that
can rapidly, efficiently and quantitatively test many chemicals. The ToxCast and
Tox21 programs are providing data from 100s of in vitro assays, on 1000s of environmental
chemicals. The first step is to estimate minimum chemical concentrations
required to significantly alter toxicity-related biological pathways in vitro, i.e.,
the Biological Pathway Altering Concentration (BPAC). Pharmacokinetic modeling
is then used to estimate the in vivo oral dose required to achieve the BPAC in
target species and tissues (e.g., human serum concentration). Uncertainties are incorporated
in both the BPAC and the pharmacokinetic parameters, and these are
combined to yield a probability distribution for the dose required to alter pathway
function. The BPAD is defined as the lower, protective end of this pathway-altering
confidence interval. We illustrate with two examples. In the first, BPADs corresponding
to the lowest AC50 for CAR/PXR-related assays were compared to the
No-Effect Level (NEL)/100 for chronic rodent liver pathology of 13 conazole
fungicides. In most cases, CAR/PXR BPADs were lower than the NEL/100, providing
a reasonable starting point for setting exposure limits if these were data poor
chemicals. In the second case, the lowest BPADs for any pathway were compared
with regulatory RfDs for 35 chemicals and, again, BPADs were mostly lower than
RfDs. BPADs were significantly higher than RfDs (less protective) in a few cases,
but most of these were cholinesterase inhibitors requiring bioactivation for toxicity.
This abstract does not necessarily reflect U.S. EPA policy.
1814 EVALUATION OF IN SILICO TOOLS FOR THE
PREDICTION OF MUTAGENICITY.
A. Hillebrecht, W. Muster, A. Brigo, M. Kansy, T. Weiser and T. Singer.
Pharmacology Research Nonclinical Safety, F. Hoffmann-La Roche Ltd., Basel,
Switzerland.
Four software packages (DEREK for Windows, Leadscope Model Applier, MCASE
MC4PC and Toxtree) were systematically evaluated for their ability to predict the
outcome of the Ames assay. A large, high quality data set of 9861 compounds, comprising
public as well as Roche proprietary data, was used to test the systems.
Acceptable performances were observed with the publicly available data sets,
whereas a significant deterioration was obtained when the assessment was performed
on the proprietary data. In particular, the sensitivity values dropped below
45 %. The relatively low amount of Ames positive compounds included in pharmaceutical
industry data sets and the different coverage of the chemical space are
the most likely explanations for the observed discrepancy between test sets. As a
general trend, expert systems (DEREK, Toxtree) tend to exhibit higher sensitivities,
but lower specificities compared to QSAR-based systems (MC4PC, Leadscope
Model Applier).