High-throughput biomonitoring for exposure discovery and ...
High-Throughput Predictive Approaches to Evaluating Chemicals in Food Contact Materials: Migration, Exposure, and Alternatives Identification Kristin Isaacs National Exposure Research Laboratory, Office of Research and Development, United States Environmental Protection Agency The views expressed in this presentation are those of the authors and do not necessarily reflect the views or policies of the U.S. EPA Forum: Identifying and Evaluating Alternative Materials: The Case of BPA-Free Can Linings UC Berkeley Center for Green Chemistry November 4, 2016 Scientific Problem The timely characterization of the human and ecological risk posed by thousands of existing and emerging commercial chemicals is a critical challenge facing EPA in its mission to protect public health and the environment
Tens of thousands chemicals in commerce have yet to be fully evaluated Example: EPAs Endocrine Disruptor Screening Program chemical list contains over 10,300 chemicals. Only 67 have undergone in vivo testing; 103 currently being tested November 29, 2014 Material from John Wambaugh 1 New High-Throughput Screening (HTS) Methods for Evaluating Hazard Tox21: Examining >10,000 chemicals using ~50 assays intended to identify interactions with biological pathways (Schmidt, 2009) EPA Toxicity Forecaster (ToxCast):
Response For a subset (>3000) of Tox21 chemicals run >1000 additional assay endpoints (Judson et al., 2010) In vitro Assay AC50 Concentration Assay AC50 with Uncertainty Concentration (mM) Material from John Wambaugh Most assays conducted in
dose-response format (identify 50% activity concentration AC50 and efficacy if data described by a Hill function) Data are being revised, new chemicals tested, new assays added All data are made public: http://comptox.epa.gov/dashboard/ 2 Exposure Predictions are Needed to Provide Real-World Context to HTS Data Near-field Consumer Products Food Far-field Household Articles Must consider multiple exposure pathways
3 Exposure Predictions are Needed to Provide Real-World Context to HTS Data Log(Exp) mg/kg/day Near-field Consumer Products Bioactive intake dose (converted with toxicokinetics as in Wetmore et al. 2015 ) Food Far-field Aggregate exposure from all relevant pathways Household Articles Must consider multiple exposure
pathways Comparing exposure to bioactivity allows for risk-based 3 prioritization Exposure Predictions are Needed to Provide Real-World Context to HTS Data Log(Exp) mg/kg/day Near-field Consumer Products Bioactive intake dose (converted with toxicokinetics as in Wetmore et al. 2015 ) Food Far-field Aggregate exposure from all
relevant pathways Household Articles Must consider multiple exposure pathways Comparing exposure to bioactivity allows for risk-based 3 prioritization Exposure Predictions are Needed to Provide Real-World Context to HTS Data Log(Exp) mg/kg/day Near-field Consumer Products Bioactive intake dose (converted with toxicokinetics as in Wetmore et al. 2015 ) Food
Far-field Aggregate exposure from all relevant pathways Household Articles Must consider multiple exposure pathways Comparing exposure to bioactivity allows for risk-based 3 prioritization Consensus Exposure Predictions with the SEEM Framework Incorporate multiple models into consensus predictions for 1000s of chemicals within the Systematic Empirical Evaluation of Models (SEEM) framework (Wambaugh et al., 2013, 2014) Evaluate/calibrate predictions with available monitoring data across as many chemical classes as possible to allow extrapolation Allows for correction of bias associated with individual models, evaluation of predictive power, and development of a consensus forecast
Evaluated by the July 2014 EPA FIFRA SAP Docket # EPA-HQ-OPP-2014-0331 4 ExpoCast Predictive Tools Relevant to Evaluation of Alternative Chemicals Having Food Contact Pathways HT models of population exposure for food contact substances (FCS) Empirical methods for predicting chemical migration Development of models and methods for screening large libraries of chemicals for functional substitutes and candidate alternatives 5 High-Throughput Predictions of Exposure from Food Contact Pathways Exposure is simply Food Concentration X Food Consumption BUTHow do we implement a HT method for ExpoCast? We identified 1659 chemicals identified as present in polymer or plastic FCS USFDAs Inventory of Effective Food Contact Substance (FCS)
Notifications and List of Indirect Additives Used in Food Contact Substances European data on plastics and surface coatings How do we approach generating migration for chemicals with different properties, from packaging with different compositions and configurations, under different storage conditions, into food substrates having different characteristics? Data-driven empirical model High uncertainty, but applicable to many chemicals 6 Prediction of Migration FDA database of migration (ug/ cm2) measurements with experimental data from polymer FCS 50 chemicals Examined steady state or maximal migration during test 1209 observations at different conditions Built linear regression model for migration level Least Absolute Shrinkage and
Selection (LASSO) algorithm to identify best subset of parameters Parameter Category FCS Properties Food Properties Food Storage Conditions Chemical Properties Parameter Description C0 Initial concentration of chemical migrant in the FCS (g/g)
Food Type Category (Food or food simulant) Alcoholic, Aqueous, Acidic, Dry, Fatty Temperature C MW S VP Molecular weight Solubility in water (mg/L) Vapor Pressure (Pa) Log (octanol/water partition coefficient) Henrys law constant (Pam3/mole) LogP HLC 7
Resulting Migration Model Most important predictors were initial concentration, vapor pressure, and temperature LogP and alcoholic food type were eliminated Vapor pressure, solubility, and LogP were correlated 8 Estimating HT Food Concentrations for Chemicals Identified in Polymer Food Contact Substances C0 distributions assigned to the 1659 FCS chemicals by using function information from data sources and migration database Migration model predictions for 15 food groups: combinations of food storage temperatures and food category Food concentration calculated using standard assumption of 6 dm2 packaging contacting 1 kg food Measured concentrations (N=276) versus maximum estimated with HT migration model
9 Predicting Population Exposures (7000 food codes) NHANES-WWEIA two-day food diaries for the years 2009-2010 (9651 diarydays) Food codes mapped to the migration model food groups Daily consumption of each food group calculated for each diary Standard FDA factors describing amount of diet contacting polymer packaging and distribution of diet across food type applied Chemicals assumed to be in all packaging (no chemical-specific prevalences) Population exposures calculated Code 11111000: Milk, cow's, fluid, whole Migration Food Group Fatty Chilled
Exposure = Concentration X Consumption 10 Comparison with Exposures Inferred from NHANES Biomarkers and Consumer Product Exposures Exposure were overestimated (as expected) given assumptions Exposures inferred from NHANES biomonitoring data as in Wambaugh et al., 2013; 2014. Biryol et al., submitted 11 Comparison with Exposures Inferred from NHANES Biomarkers and Consumer Product Exposures Exposure were overestimated (as expected) given assumptions Exposures inferred from NHANES biomonitoring data as in Wambaugh et al. Environmental Science and Technology, 2014.
but correlated with inferences and also associated with inferences when included in an initial aggregate model with consumer product exposures Biryol et al., submitted 11 Databases and Models for Predicting Function of Chemicals Plasticizer Colorant Emulsifier Defoamer Chelating Agent UV absorber Surfactant
Solvent Functional Use (FUse) Dataset 14,000+ Chemicals 200+ Functions Fragrance Fill gaps in quantitative information Refined heuristics of exposure Other applications Allows for Modeling of Function in Terms of Chemical Properties or Structures 12 Isaacs et al., 2016 Classification Models for Chemical Function Chemical Structure and Property Descriptors
Chemical Function Information FUSE Prediction of Of Potential Functional Substitutes EPI-SuiteTM Machine-Learning Based Classification 13 Classification Models for Chemical Function Chemical Structure and Property Descriptors Chemical Function Information FUSE Prediction of Of Potential Functional Substitutes
EPI-Suite TM We have good predictive models for a number of FCS-related functions Machine-Learning Based Classification 13 Application: Screening for Alternatives By Function and Bioactivity Phillips et al., submitted 14 Application: Screening for Alternatives By Function and Bioactivity Comparing a metric of bioactivity (across a number of Tox21 assays) for predicted functional substitutes against a threshold value derived from existing chemicals
with that function identified 648 candidate alternatives Phillips et al., submitted 15 A New Source of Curated HT Chemical Information http://comptox.epa.gov/dashboard/ Data on over 700,000 chemicals Property predictions when QSAR models available New curated structural information being incorporated Will allow us to expand the prediction of functional substitutes to larger libraries of chemicals Technical leads Tony Williams, Richard Judson, et al. (NCCT) 16 Conclusions
US EPA is generating high-throughput predictive models for use in riskbased prioritization of chemicals for further study While our HT migration and exposure predictions by their nature have higher uncertainty than focused single-chemical assessments, the data and approaches could inform evaluation of alternatives when no other information is available HT dietary exposure framework could incorporate refined data that could improve exposure estimates (e.g. refined composition information, chemical occurrence in packaging, even measured migration or concentration data) Aggregate predictions from ExpoCast can also inform background exposures for proposed alternatives (i.e. from sources other than FCS applications) We are working to expand the application of classification models for functional use to identify previously unknown compounds that could be further evaluated as alternatives for existing chemicals 17 Collaborators Chemical Safety for Sustainability (CSS) Rapid Exposure and Dosimetry (RED) Project NCCT Chris Grulke Richard Judson Dustin Kapraun* Andrew McEachran* Robert Pearce*
Ann Richard Risa Sayre Woody Setzer Rusty Thomas NRMRL Yirui Liang* Xiaoyu Liu NHEERL Linda Adams Christopher Ecklund Marina Evans Mike Hughes Jane Ellen Simmons John Wambaugh Antony Williams *Trainees Lead CSS Matrix Interface: John Kenneke (NERL) NERL Craig Barber Derya Biryol* Namdi Brandon*
Kathie Dionisio Peter Egeghy Brandall Ingle* Kristin Isaacs Seth Newton Katherine Phillips Paul Price Mark Strynar Jon Sobus Cecilia Tan Mike Tornero-Velez Elin Ulrich Dan Vallero Barbara Wetmore Arnot Research and Consulting Jon Arnot Battelle Memorial Institute Anne Louise Sumner Anne Gregg Chemical Computing Group Rocky Goldsmith National Institute for Environmental Health Sciences (NIEHS) National Toxicology Program Mike Devito
Steve Ferguson Nisha Sipes Netherlands Organisation for Applied Scientific Research (TNO) Sieto Bosgra North Carolina Central University Chantel Nicolas Research Triangle Institute Timothy Fennell ScitoVation Harvey Clewell Silent Spring Institute Robin Dodson Southwest Research Institute Alice Yau Kristin Favela Summit Toxicology Lesa Aylward Tox Strategies Caroline Ring University of California, Davis Deborah Bennett Hyeong-Moo Shin University of Michigan Olivier Jolliet University of North Carolina, Chapel Hill
Alex Tropsha The views expressed in this presentation are those of the author and do not necessarily reflect the views or policies of the U.S. EPA References Biryol, Derya, et al. High-Throughput Dietary Exposure Predictions for Chemical Migrants from Food Contact Substances for Use in Chemical Prioritization, submitted Isaacs, Kristin K., et al. "Characterization and Prediction of Chemical Functions and Weight Fractions in Consumer Products. Toxicology Reports (2016) Wambaugh, John F., et al. "High Throughput Heuristics for Prioritizing Human Exposure to Environmental Chemicals." Environmental science & technology (2014). Phillips, Katherine et al. High-throughput Screening of Chemicals as Functional Substitutes using Structure-based Classification Models. Submitted. Wetmore, Barbara A., et al. "Incorporating High-Throughput Exposure Predictions with Dosimetry-Adjusted In Vitro Bioactivity to Inform Chemical Toxicity Testing." Toxicological Sciences 148.1 (2015): 121-136.
Linux Password Vault. This project aims to create a corporate password repository where all corporate passwords can be stored securely, accessed by the authorized personnel only on a need-to-know basis.
Edgar Degas Shape 2D: The area enclosed by an outline SHAPE "When the subject is strong, simplicity is the only way to treat i" ""Jacob Lawrence Form 3D: height, width and depth FORM "There is no retirement for an artist,...
A Gated Latch. Adding a second level of AND gates to a SR latch. Enable signal. When . gate . signal = 0 (disable), the latch remains unchanged (hold). When . gate. signal = 1 (enable), the latch acts at...
Smart Medical Home Israel Gannot Bio-Medical Engineering Department, School of Engineering, Tel-Aviv University George Washington University Washington, DC Introduction The demographic trends of the advanced countries go towards the development of an older population.
POC = Immunochromatography. Antibodies are made in such a way that they bind with a specific drug, such as morphine. In one approach, manufacturers of point of care (POC) devices embed test strips with antibodies and install them in devices...
Ready to download the document? Go ahead and hit continue!