Christopher M. Grulke

CERAPP : Collaborative Estrogen Receptor Activity Prediction Project

Background: Humans are exposed to thousands of man-made chemicals in the environment. Some chemicals mimic natural endocrine hormones and, thus, have the potential to be endocrine disruptors. Most of these chemicals have never been tested for their ability to interact with the estrogen receptor (ER). Risk assessors need tools to prioritize chemicals for evaluation in costly in vivo tests, for instance, within the U.S. EPA Endocrine Disruptor Screening Program. Objectives: We describe a large-scale modeling project called CERAPP (Collaborative Estrogen Receptor Activity Prediction Project) and demonstrate the efficacy of using predictive computational models trained on high-throughput screening data to evaluate thousands of chemicals for ER-related activity and prioritize them for further testing. Methods: CERAPP combined multiple models developed in collaboration with 17 groups in the United States and Europe to predict ER activity of a common set of 32,464 chemical structures. Quantitative structure–activity relationship models and docking approaches were employed, mostly using a common training set of 1,677 chemical structures provided by the U.S. EPA, to build a total of 40 categorical and 8 continuous models for binding, agonist, and antagonist ER activity. All predictions were evaluated on a set of 7,522 chemicals curated from the literature. To overcome the limitations of single models, a consensus was built by weighting models on scores based on their evaluated accuracies. Results: Individual model scores ranged from 0.69 to 0.85, showing high prediction reliabilities. Out of the 32,464 chemicals, the consensus model predicted 4,001 chemicals (12.3%) as high priority actives and 6,742 potential actives (20.8%) to be considered for further testing. Conclusion: This project demonstrated the possibility to screen large libraries of chemicals using a consensus of different in silico approaches. This concept will be applied in future projects related to other end points. Citation: Mansouri K, Abdelaziz A, Rybacka A, Roncaglioni A, Tropsha A, Varnek A, Zakharov A, Worth A, Richard AM, Grulke CM, Trisciuzzi D, Fourches D, Horvath D, Benfenati E, Muratov E, Wedebye EB, Grisoni F, Mangiatordi GF, Incisivo GM, Hong H, Ng HW, Tetko IV, Balabin I, Kancherla J, Shen J, Burton J, Nicklaus M, Cassotti M, Nikolov NG, Nicolotti O, Andersson PL, Zang Q, Politi R, Beger RD, Todeschini R, Huang R, Farag S, Rosenberg SA, Slavov S, Hu X, Judson RS. 2016. CERAPP: Collaborative Estrogen Receptor Activity Prediction Project. Environ Health Perspect 124:1023–1033; http://dx.doi.org/10.1289/ehp.1510267

Linking high resolution mass spectrometry data with exposure and toxicity forecasts to advance high-throughput environmental monitoring

There is a growing need in the field of exposure science for monitoring methods that rapidly screen environmental media for suspect contaminants. Measurement and analysis platforms, based on high resolution mass spectrometry (HRMS), now exist to meet this need. Here we describe results of a study that links HRMS data with exposure predictions from the U.S. EPAs ExpoCast™ program and in vitro bioassay data from the U.S. interagency Tox21 consortium. Vacuum dust samples were collected from 56 households across the U.S. as part of the American Healthy Homes Survey (AHHS). Sample extracts were analyzed using liquid chromatography time-of-flight mass spectrometry (LC-TOF/MS) with electrospray ionization. On average, approximately 2000 molecular features were identified per sample (based on accurate mass) in negative ion mode, and 3000 in positive ion mode. Exact mass, isotope distribution, and isotope spacing were used to match molecular features with a unique listing of chemical formulas extracted from EPAs Distributed Structure-Searchable Toxicity (DSSTox) database. A total of 978 DSSTox formulas were consistent with the dust LC-TOF/molecular feature data (match score≥90); these formulas mapped to 3228 possible chemicals in the database. Correct assignment of a unique chemical to a given formula required additional validation steps. Each suspect chemical was prioritized for follow-up confirmation using abundance and detection frequency results, along with exposure and bioactivity estimates from ExpoCast and Tox21, respectively. Chemicals with elevated exposure and/or toxicity potential were further examined using a mixture of 100 chemical standards. A total of 33 chemicals were confirmed present in the dust samples by formula and retention time match; nearly half of these do not appear to have been associated with house dust in the published literature. Chemical matches found in at least 10 of the 56 dust samples include Piperine, N,N-Diethyl-m-toluamide (DEET), Triclocarban, Diethyl phthalate (DEP), Propylparaben, Methylparaben, Tris(1,3-dichloro-2-propyl)phosphate (TDCPP), and Nicotine. This study demonstrates a novel suspect screening methodology to prioritize chemicals of interest for subsequent targeted analysis. The methods described here rely on strategic integration of available public resources and should be considered in future non-targeted and suspect screening assessments of environmental and biological media.

SHEDS-HT: an integrated probabilistic exposure model for prioritizing exposures to chemicals with near-field and dietary sources.

United States Environmental Protection Agency (USEPA) researchers are developing a strategy for high-throughput (HT) exposure-based prioritization of chemicals under the ExpoCast program. These novel modeling approaches for evaluating chemicals based on their potential for biologically relevant human exposures will inform toxicity testing and prioritization for chemical risk assessment. Based on probabilistic methods and algorithms developed for The Stochastic Human Exposure and Dose Simulation Model for Multimedia, Multipathway Chemicals (SHEDS-MM), a new mechanistic modeling approach has been developed to accommodate high-throughput (HT) assessment of exposure potential. In this SHEDS-HT model, the residential and dietary modules of SHEDS-MM have been operationally modified to reduce the user burden, input data demands, and run times of the higher-tier model, while maintaining critical features and inputs that influence exposure. The model has been implemented in R; the modeling framework links chemicals to consumer product categories or food groups (and thus exposure scenarios) to predict HT exposures and intake doses. Initially, SHEDS-HT has been applied to 2507 organic chemicals associated with consumer products and agricultural pesticides. These evaluations employ data from recent USEPA efforts to characterize usage (prevalence, frequency, and magnitude), chemical composition, and exposure scenarios for a wide range of consumer products. In modeling indirect exposures from near-field sources, SHEDS-HT employs a fugacity-based module to estimate concentrations in indoor environmental media. The concentration estimates, along with relevant exposure factors and human activity data, are then used by the model to rapidly generate probabilistic population distributions of near-field indirect exposures via dermal, nondietary ingestion, and inhalation pathways. Pathway-specific estimates of near-field direct exposures from consumer products are also modeled. Population dietary exposures for a variety of chemicals found in foods are combined with the corresponding chemical-specific near-field exposure predictions to produce aggregate population exposure estimates. The estimated intake dose rates (mg/kg/day) for the 2507 chemical case-study spanned 13 orders of magnitude. SHEDS-HT successfully reproduced the pathway-specific exposure results of the higher-tier SHEDS-MM for a case-study pesticide and produced median intake doses significantly correlated (p<0.0001, R2=0.39) with medians inferred using biomonitoring data for 39 chemicals from the National Health and Nutrition Examination Survey (NHANES). Based on the favorable performance of SHEDS-HT with respect to these initial evaluations, we believe this new tool will be useful for HT prediction of chemical exposure potential.

Chembench: a cheminformatics workbench

MOTIVATION Advances in the field of cheminformatics have been hindered by a lack of freely available tools. We have created Chembench, a publicly available cheminformatics portal for analyzing experimental chemical structure-activity data. Chembench provides a broad range of tools for data visualization and embeds a rigorous workflow for creating and validating predictive Quantitative Structure-Activity Relationship models and using them for virtual screening of chemical libraries to prioritize the compound selection for drug discovery and/or chemical safety assessment. AVAILABILITY Freely accessible at: http://chembench.mml.unc.edu CONTACT alex_tropsha@unc.edu

An automated curation procedure for addressing chemical errors and inconsistencies in public datasets used in QSAR modelling

Abstract The increasing availability of large collections of chemical structures and associated experimental data provides an opportunity to build robust QSAR models for applications in different fields. One common concern is the quality of both the chemical structure information and associated experimental data. Here we describe the development of an automated KNIME workflow to curate and correct errors in the structure and identity of chemicals using the publicly available PHYSPROP physicochemical properties and environmental fate datasets. The workflow first assembles structure–identity pairs using up to four provided chemical identifiers, including chemical name, CASRNs, SMILES, and MolBlock. Problems detected included errors and mismatches in chemical structure formats, identifiers and various structure validation issues, including hypervalency and stereochemistry descriptions. Subsequently, a machine learning procedure was applied to evaluate the impact of this curation process. The performance of QSAR models built on only the highest-quality subset of the original dataset was compared with the larger curated and corrected dataset. The latter showed statistically improved predictive performance. The final workflow was used to curate the full list of PHYSPROP datasets, and is being made publicly available for further usage and integration by the scientific community.

Development, Validation, and Use of Quantitative Structure−Activity Relationship Models of 5-Hydroxytryptamine (2B) Receptor Ligands to Identify Novel Receptor Binders and Putative Valvulopathic Compounds among Common Drugs

Some antipsychotic drugs are known to cause valvular heart disease by activating serotonin 5-HT(2B) receptors. We have developed and validated binary classification QSAR models capable of predicting potential 5-HT(2B) actives. The classification accuracies of the models built to discriminate 5-HT(2B) actives from the inactives were as high as 80% for the external test set. These models were used to screen in silico 59,000 compounds included in the World Drug Index, and 122 compounds were predicted as actives with high confidence. Ten of them were tested in radioligand binding assays and nine were found active, suggesting a success rate of 90%. All validated actives were then tested in functional assays, and one compound was identified as a true 5-HT(2B) agonist. We suggest that the QSAR models developed in this study could be used as reliable predictors to flag drug candidates that are likely to cause valvulopathy.

Prediction of Aquatic Toxicity Mode of Action Using Linear Discriminant and Random Forest Models

The ability to determine the mode of action (MOA) for a diverse group of chemicals is a critical part of ecological risk assessment and chemical regulation. However, existing MOA assignment approaches in ecotoxicology have been limited to a relatively few MOAs, have high uncertainty, or rely on professional judgment. In this study, machine based learning algorithms (linear discriminant analysis and random forest) were used to develop models for assigning aquatic toxicity MOA. These methods were selected since they have been shown to be able to correlate diverse data sets and provide an indication of the most important descriptors. A data set of MOA assignments for 924 chemicals was developed using a combination of high confidence assignments, international consensus classifications, ASTER (ASessment Tools for the Evaluation of Risk) predictions, and weight of evidence professional judgment based an assessment of structure and literature information. The overall data set was randomly divided into a training set (75%) and a validation set (25%) and then used to develop linear discriminant analysis (LDA) and random forest (RF) MOA assignment models. The LDA and RF models had high internal concordance and specificity and were able to produce overall prediction accuracies ranging from 84.5 to 87.7% for the validation set. These results demonstrate that computational chemistry approaches can be used to determine the acute toxicity MOAs across a large range of structures and mechanisms.

A Workflow to Investigate Exposure and Pharmacokinetic Influences on High-Throughput in Vitro Chemical Screening Based on Adverse Outcome Pathways

Background Adverse outcome pathways (AOPs) link adverse effects in individuals or populations to a molecular initiating event (MIE) that can be quantified using in vitro methods. Practical application of AOPs in chemical-specific risk assessment requires incorporation of knowledge on exposure, along with absorption, distribution, metabolism, and excretion (ADME) properties of chemicals. Objectives We developed a conceptual workflow to examine exposure and ADME properties in relation to an MIE. The utility of this workflow was evaluated using a previously established AOP, acetylcholinesterase (AChE) inhibition. Methods Thirty chemicals found to inhibit human AChE in the ToxCast™ assay were examined with respect to their exposure, absorption potential, and ability to cross the blood–brain barrier (BBB). Structures of active chemicals were compared against structures of 1,029 inactive chemicals to detect possible parent compounds that might have active metabolites. Results Application of the workflow screened 10 “low-priority” chemicals of 30 active chemicals. Fifty-two of the 1,029 inactive chemicals exhibited a similarity threshold of ≥ 75% with their nearest active neighbors. Of these 52 compounds, 30 were excluded due to poor absorption or distribution. The remaining 22 compounds may inhibit AChE in vivo either directly or as a result of metabolic activation. Conclusions The incorporation of exposure and ADME properties into the conceptual workflow eliminated 10 “low-priority” chemicals that may otherwise have undergone additional, resource-consuming analyses. Our workflow also increased confidence in interpretation of in vitro results by identifying possible “false negatives.” Citation Phillips MB, Leonard JA, Grulke CM, Chang DT, Edwards SW, Brooks R, Goldsmith MR, El-Masri H, Tan YM. 2016. A workflow to investigate exposure and pharmacokinetic influences on high-throughput in vitro chemical screening based on adverse outcome pathways. Environ Health Perspect 124:53–60; http://dx.doi.org/10.1289/ehp.1409450

Integrating tools for non-targeted analysis research and chemical safety evaluations at the US EPA

Tens-of-thousands of chemicals are registered in the U.S. for use in countless processes and products. Recent evidence suggests that many of these chemicals are measureable in environmental and/or biological systems, indicating the potential for widespread exposures. Traditional public health research tools, including in vivo studies and targeted analytical chemistry methods, have been unable to meet the needs of screening programs designed to evaluate chemical safety. As such, new tools have been developed to enable rapid assessment of potentially harmful chemical exposures and their attendant biological responses. One group of tools, known as “non-targeted analysis” (NTA) methods, allows the rapid characterization of thousands of never-before-studied compounds in a wide variety of environmental, residential, and biological media. This article discusses current applications of NTA methods, challenges to their effective use in chemical screening studies, and ways in which shared resources (e.g., chemical standards, databases, model predictions, and media measurements) can advance their use in risk-based chemical prioritization. A brief review is provided of resources and projects within EPA’s Office of Research and Development (ORD) that provide benefit to, and receive benefits from, NTA research endeavors. A summary of EPA’s Non-Targeted Analysis Collaborative Trial (ENTACT) is also given, which makes direct use of ORD resources to benefit the global NTA research community. Finally, a research framework is described that shows how NTA methods will bridge chemical prioritization efforts within ORD. This framework exists as a guide for institutions seeking to understand the complexity of chemical exposures, and the impact of these exposures on living systems.

Developing a Physiologically-Based Pharmacokinetic Model Knowledgebase in Support of Provisional Model Construction

Developing physiologically-based pharmacokinetic (PBPK) models for chemicals can be resource-intensive, as neither chemical-specific parameters nor in vivo pharmacokinetic data are easily available for model construction. Previously developed, well-parameterized, and thoroughly-vetted models can be a great resource for the construction of models pertaining to new chemicals. A PBPK knowledgebase was compiled and developed from existing PBPK-related articles and used to develop new models. From 2,039 PBPK-related articles published between 1977 and 2013, 307 unique chemicals were identified for use as the basis of our knowledgebase. Keywords related to species, gender, developmental stages, and organs were analyzed from the articles within the PBPK knowledgebase. A correlation matrix of the 307 chemicals in the PBPK knowledgebase was calculated based on pharmacokinetic-relevant molecular descriptors. Chemicals in the PBPK knowledgebase were ranked based on their correlation toward ethylbenzene and gefitinib. Next, multiple chemicals were selected to represent exact matches, close analogues, or non-analogues of the target case study chemicals. Parameters, equations, or experimental data relevant to existing models for these chemicals and their analogues were used to construct new models, and model predictions were compared to observed values. This compiled knowledgebase provides a chemical structure-based approach for identifying PBPK models relevant to other chemical entities. Using suitable correlation metrics, we demonstrated that models of chemical analogues in the PBPK knowledgebase can guide the construction of PBPK models for other chemicals.