The Toxicologist - Society of Toxicology
The Toxicologist - Society of Toxicology
The Toxicologist - Society of Toxicology
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
suggest that the combination <strong>of</strong> tools to address reactive metabolite risk can provide<br />
design teams with an assessment <strong>of</strong> real risk rather than the simple hazard alert <strong>of</strong> a<br />
structural alert.<br />
155 DESIGN AND DEVELOPMENT OF AN<br />
INSTITUTIONAL KNOWLEDGE-BASE AT FDA’S<br />
CENTER FOR FOOD SAFETY AND APPLIED<br />
NUTRITION.<br />
K. Arvidson, A. McCarthy, C. Yang and D. Hristozov. CFSAN, U.S. FDA,<br />
College Park, MD.<br />
<strong>The</strong> Chemical Evaluation and Risk Estimation System (CERES) project under development<br />
in FDA’s Center for Food Safety and Applied Nutrition aims at establishing<br />
a sustainable data management and storage system that will provide decision<br />
support tools for both pre-market and post-market safety assessments <strong>of</strong> food additives<br />
and food contact substances as well as for potential contamination issues. <strong>The</strong><br />
development <strong>of</strong> CERES will provide a single unified data repository that compiles<br />
available information on a substance, including: chemical structures and properties,<br />
regulation records, toxicity studies, and other biological screening assays. In cases<br />
where no information is available for a particular substance, CERES provides tools<br />
to identify potential safety concerns by applying mode <strong>of</strong> action-driven QSAR prediction<br />
models as well as to identify and analyze data on structural and biological<br />
analogs (read-across). To achieve this goal, CERES requires a solid foundation for<br />
handling chemical and toxicity data as well as knowledge (e.g., rules or computational<br />
toxicology models) derived from these data. Thus the CERES project not<br />
only needs to address a wide variety <strong>of</strong> issues in database and knowledgebase construction,<br />
but also in the development <strong>of</strong> computational methods. Strategies are<br />
presented for tasks <strong>of</strong> data harvesting, data modeling, development <strong>of</strong> rule-base and<br />
computational models. This poster will focus on the overall strategy for the CERES<br />
knowledgebase; each sub-area will also be presented in other posters.<br />
156 PHARMACOLOGICAL PROFILE SIMILARITY – A<br />
RELEVANT METHOD FOR ANTICIPATION OF IN VIVO<br />
SAFETY?<br />
D. Muthas 1 , M. Rolf 2 , S. Matis-Mitchell 3 and S. Boyer 1 . 1 Global Safety<br />
Assessment, AstraZeneca R&D, Mölndal, Sweden, 2 Global Safety Assessment,<br />
AstraZeneca R&D, Alderly Park, United Kingdom and 3 Discovery Information,<br />
AstraZeneca R&D, Wilmington, DE.<br />
We have analyzed a pharmacological pr<strong>of</strong>ile <strong>of</strong> historical AstraZeneca development<br />
compounds in order to investigate their potential to predict in vivo findings. A<br />
panel <strong>of</strong> 100 targets, most <strong>of</strong> which have been chosen for their association with adverse<br />
effects, was used to derive a pharmacological ‘biospectrum’ <strong>of</strong> each compound.<br />
<strong>The</strong> similarities <strong>of</strong> these biospectra were then calculated using several different<br />
similarity metrics (e.g. Tanimoto, Dice, Yule, Euclidian distance and<br />
Mahalanobis distance) and activity binning to identify compounds sharing similar<br />
biological fingerprints. To evaluate their relevance, these similarities were then compared<br />
to the preclinical and/or clinical findings <strong>of</strong> the compounds found by mining<br />
preclinical study reports and clinical safety databases. <strong>The</strong> observations were later<br />
analyzed using all found data as well as with special focus on certain types <strong>of</strong> studies<br />
(eg 1 month dog and 2 week rat). <strong>The</strong> compounds were then clustered based in<br />
the in vivo effects and compared to the clusters based on biospectrum. <strong>The</strong> results<br />
<strong>of</strong> these studies indicate that several pairs <strong>of</strong> structurally dissimilar compounds with<br />
similar biospectra can be identified and an overlap <strong>of</strong> their preclinical findings was<br />
frequently found. <strong>The</strong> frequency <strong>of</strong> association <strong>of</strong> the biospectrum and outcome<br />
was assessed for statistical significance and found to be significant in several cases.<br />
This suggests that pharmacological pr<strong>of</strong>iling and establishing a biospectrum could<br />
help identify the potential for safety issues from more than the chemical structure<br />
or the activity on a single target.<br />
157 DEVELOPMENT OF QSAR MODEL FOR HIGH-SPEED<br />
IN SILICO IDENTIFICATION OF POTENTIALLY<br />
PHOTOTOXIC ORGANIC COMPOUNDS.<br />
W. Plonka 1 and J. M. Ciloy 2 . 1 Life Science, FQS Poland, Cracow, Poland and 2 Life<br />
Science, Fujitsu Kyushu Systems Ltd., Fukuoka, Japan. Sponsor: K. Hayamizu.<br />
Light-exposure induced toxicity to humans is one <strong>of</strong> factors limiting the use <strong>of</strong><br />
chemical compounds in wide range <strong>of</strong> applications, starting with cosmetics, household<br />
chemistry products, as well as pharmaceutical industry. <strong>The</strong> possibility <strong>of</strong> reliable<br />
in-silico identification <strong>of</strong> potentially phototoxic compounds would contribute<br />
to product safety and reduce development costs. A Qualitative Structure-Activity<br />
Relationship study had been done on a set <strong>of</strong> 229 structures, <strong>of</strong> which 114 were<br />
known from public literature to cause phototoxic effects, and the rest were neutral<br />
compounds, including compounds routinely used in commercial cosmetics. <strong>The</strong><br />
objective was to create a model that would have as low as possible false-negative<br />
classification rate to ensure high safety margin <strong>of</strong> the predictions. It was assumed,<br />
that a molecule has to satisfy a set <strong>of</strong> quantum, as well as structural properties – to<br />
which both light absorption and energy transfer in biological systems relate in order<br />
to be potentially phototoxic. A stochastic gradient perceptron model was created<br />
employing quantum descriptors calculated by fast AM1 semiempirical method,<br />
fragment count and topological descriptors. <strong>The</strong> model has a 0% false-negative<br />
classification rate on the training set. <strong>The</strong> overall internal leave-1-out cross validation<br />
rate <strong>of</strong> the model is over 90%. <strong>The</strong> details <strong>of</strong> model building technique and<br />
compounds used in research are presented in the poster.<br />
158 THE POWER OF AN ONTOLOGY-DRIVEN<br />
DEVELOPMENTAL TOXICITY DATABASE FOR DATA<br />
MINING AND COMPUTATIONAL MODELING.<br />
A. Richard 1 , E. Julien 2 , E. Busta 3 , M. Wolf 4 and C. Yang 3 . 1 NCCT (D343-03),<br />
U.S. EPA, Research Triangle Park, NC, 2 LifeSci Research Service, Rockville, MD,<br />
3 HFS-2775, U.S. FDA CFSA, College Park, Maryland, MD and 4 Lockheed-Martin,<br />
U.S. EPA, Research Triangle Park, NC.<br />
Modeling <strong>of</strong> developmental toxicology presents a significant challenge to computational<br />
toxicology due to endpoint complexity and lack <strong>of</strong> data coverage. <strong>The</strong>se challenges<br />
largely account for the relatively few modeling successes using the structure–activity<br />
relationship (SAR) paradigm. <strong>The</strong> development <strong>of</strong> new in vitro<br />
pr<strong>of</strong>iling approaches, employing screening assays or lower organisms to evaluate developmental<br />
toxicity requires anchoring to the results <strong>of</strong> in vivo studies. <strong>The</strong><br />
International Life Science Institute (ILSI) recently released a public ontologydriven<br />
DevToxDB with data extracted from published studies. This project heavily<br />
influenced other efforts, and the data content in the ILSI DevToxDB was combined<br />
into an ontology-based database along with other public data from EPA<br />
ToxRefDB, FDA Center for Food Safety and Applied Nutrition and Center for<br />
Drug Evaluation Research and the National <strong>Toxicology</strong> Program. All chemical<br />
structures are indexed in DSSTox, providing the ability to assess chemical coverage<br />
and diversity <strong>of</strong> these largely non-overlapping data inventories, which include<br />
drugs, pesticides, industrial chemicals, food ingredients and contact substances.<br />
<strong>The</strong> use <strong>of</strong> a common toxicological ontology provides a logical means to group and<br />
aggregate biological effects in subsequent computational analyses.<br />
Chemoinformatics methods are used to compare the chemical space <strong>of</strong> each data<br />
source, and to explore associations with biological endpoints through the toxicology<br />
ontology perspective. Chemicals in these distinct datasets cause common phenotypes<br />
as well as distinct, non-overlapping phenotypes in the same target organs.<br />
Incorporation <strong>of</strong> the ILSI DevTox database into other public database efforts<br />
should provide a rich foundation for spurring innovation in SAR modeling <strong>of</strong> developmental<br />
endpoints. This abstract does not necessarily represent policies <strong>of</strong> FDA<br />
or EPA.<br />
159 USE OF C. ELEGANS GROWTH ASSAY DATA IN NOVEL<br />
TWO-STEP HIERARCHICAL QSAR MODELING<br />
WORKFLOW ENHANCES IN VIVO TOXICITY<br />
PREDICTION FOR TOXCAST PHASE I CHEMICALS.<br />
A. Golbraikh 1, 2 , R. Shah 1, 4 , W. Boyd 3 , M. Smith 4 , A. Tropsha 1, 2 and J.<br />
Freedman 3 . 1 SciOme LLC, Research Triangle Park, NC, 2 University <strong>of</strong> North<br />
Carolina at Chapel Hill, Chapel Hill, NC, 3 National Institute <strong>of</strong> Environmental<br />
Health Science, Research Triangle Park, NC and 4 SRA International, Durham, NC.<br />
<strong>The</strong> US EPA ToxCast Phase I program provides data for 320 substances with<br />
known in vivo toxicity measured in 76 assays; the results <strong>of</strong> ca. 600 in vitro assays<br />
for the same substances are available as well. <strong>The</strong> latter data have been used previously<br />
with varying degree <strong>of</strong> success to improve the predictive power <strong>of</strong> in silico<br />
models <strong>of</strong> chemical toxicity. Herein, we have explored, in the same context, new<br />
whole organism toxicity data generated for ToxCast chemicals in the C. elegans<br />
growth assay. <strong>The</strong> goal <strong>of</strong> this study was to establish both the absolute as well as relative<br />
(i.e., with respect to other short term assays included in the ToxCast Phase I<br />
database) value <strong>of</strong> C. elegans growth assay for predicting chemical in vivo toxicity<br />
and prioritizing new chemicals for in vivo studies. Unlike most in vitro assays where<br />
the number <strong>of</strong> active compounds was typically lower than that <strong>of</strong> inactive ones, the<br />
C. elegans assay resulted in nearly balanced dataset. kNN QSAR modeling <strong>of</strong> C. elegans<br />
data using our standard computational workflow with the emphasis on external<br />
validation resulted in models with the total accuracy <strong>of</strong> 66%. <strong>The</strong> C. elegans<br />
SOT 2011 ANNUAL MEETING 33