topical scientific workshop on new approach methodologies...
TRANSCRIPT
• Topical Scientific Workshop on New Approach Methodologies in Regulatory Science
19-20 April 2016
Dr Tomasz Sobański
European Chemicals Agency,Finland
Feedback
• Paper forms available on the side table: online survey link on the workshop website
4
Reports from break-outsessions
Rapporteurs:
Dr Norbert Fedtke, ECHA
Dr Elisabet Berggren, European Commission
Dr Karel de Raat, advisor to ECHA
Break-out session 1
Case study from SEURAT-1
Perfluorinated alkyl acids: direct acting toxicant category supported by ToxCast evidence
Chair:
Dr Watze de Wolf, Chairman of the Member State Committee, ECHA, Finland
Presenter:
Ms Sharon Stuard, Procter & Gamble,
United States of America
Rapporteur:
Dr Norbert Fedtke, ECHA, Finland
Topical Scientific Workshop on
New Approach Methodologies in Regulatory Science
19 – 20 April, 2016
4/20/2016 7
Question block 1: Case description
1. Are the chemical structures from the source substances and target substances sufficiently described to determine the structural (dis)similarities?
Yes, impurity profiles would be needed though (no further consideration for the current case)
2. What is the regulatory purpose for which the case is prepared?
Filling data gap for 90-day RDT in the context of REACH registration
3. What is the property for which a prediction is attempted?
90-day RDT with the critical effect liver toxicity with the purpose to identify a NOAEL and subsequently derive a DNEL
4. What is the hypothesis under which the prediction is attempted?
Parent substance is the acting chemical structure, no metabolism. Liver toxicity as key effect mediated via receptor interaction (PPAR). Variation in strength of effects
Case description (2)
• 5. What information obtained in animal studies is provided to support the prediction?
90-day study with source substance, “flanking information” on C 6, C11, C12 supporting the prediction, but not in the category.
• 6. What are the weak points of the justification, which are identified by the assessment according to the RAAF?
• Borders of the category; justification could be improved for the selection of category members.
• C8 selected as worst case for liver toxicity. Really worst case for the set of parameters in a 90-day study? Issues are: toxicokinetic information and possible other effects, e.g. thyroid function disruption, haematological, clear statement missing on effect doses for those possible effects in vivo
• Documentation of the weight of the mechanistic evidence.
•4/20/2016 •8
• 7. What supporting evidence is still missing or could be added to increase the confidence in the prediction?
• Data table for ToxCast, other NAM data; further toxicokinetic information
• Scientifically relevant question, regulatory-wise obligation met without NAM
• Systematic assessment of quality of all information provided
• Further detailed description of the methods used in Toxcast and of the results obtained
•4/20/2016 •9
4/20/2016 10
Question block 2: Contribution of NAM information
8. What types of NAM information have been used in the case to increase the confidence in the prediction?
Mechanistic understanding of the liver toxicity, ToxCast assays
9. Were the weak points in the prediction addressed or at least partly addressed by NAM, and has the condifence in the prediction increased
Lacking further toxicokinetic information remains a weak point, NAM data currently in the case study not supposed to address this.
4/20/2016 11
Question block 3: Detailed characteristics of the NAM information provided
10. Are the NAM methods used in the case study generally available?
no
11. What scientific limitations are evident for the NAM information in the case study?
Solubility issues of test substances, applicability domain due to phys-chemissues
12. What could be the barriers in using or generating the necessary NAM data?
Cost currently, expected to decrease in importance in future , time, uncertainty on acceptance (from the panel discussion), missing reporting standards, potential for predicting absence of effect
Break-out session 2
Case study from SEURAT-1
β-Unsaturated alcohols: indirect acting toxicant category supported by SEURAT-1
Chair: Dr Derek Knight, Senior Scientific Advisor, ECHA, Finland
Presenter: Dr Andrea Richarz, European Commission, Joint Research Centre, Italy
Rapporteur: Dr Elisabet Berggren, European Commission, Joint Research Centre, Italy
Topical Scientific Workshop on
New Approach Methodologies in Regulatory Science
19 – 20 April, 2016
Major Uncertainties
• Complexity of structures → Similarity / category boundaries and members
• Details on study design / quality of in vivo data, choice of NOAEL
• Potency of effects, order of reactivity
→ proving the worst case for source compound
• Transformation mechanism (other than via ADH)/rates?
→ reactive potency vs kinetics (size of the category)
• Variation of metabolic pathways (aldehydes/ketones)
→ possible other effects via further metabolitespresent (kinetics of transformations)
• Toxic reactivity mechanism? Map on AOP?
Read-across issues discussed:
• Toxicokinetics is challenging also in terms of what data is needed to support a given hypothesis.
• How to identify the category, in reality depending on interest from the manufacturer.
4/20/2016 14INTERNAL
Information Added and Uncertainties Reduced by NAM
• Metabolism of βOAs to toxicant, reactivity of metabolites as opposed to parent compounds, link with structure (ex vivo perfused liver, in silico, in chemico)
• in chemico data: only quantitative data available to show clustering of reactivity potency (supported by in silico/in vitro)
• Mechanism of adverse effect evidence strengthened by in chemico/in silico/ SEURAT-1 in vitro data
• in particular: HSC activation markers in hepatic organoids confirm MoA hypothesis of metabolic-mediated fibrosis
How can NAM/non-animal data support read-across? (1)
• NAM is providing mechanistic information that together with other information, e.g. in vivo data, assist in establishing the hypothesis.
• Evaluate whether the NAM data is relevant to the specific read-across case.
• To split the case into a two-step procedure:
o First addressing metabolic rate, e.g. metabolic profile in human hepatocytes or HepRG.
o Secondly the reactivity, e.g. evidence could be molecular mechanism
4/20/2016 16INTERNAL
How can NAM/non-animal data support read-across? (2)
• Be creative using different evidence, e.g. skin sensitization can be sued to help inform on chemical reactivity and biological availability and activation.
• In vivo omics data could assist to bridge in between the in vitro and in vivo situation.
• Evidence from non vertebrate species, knowing which biological pathways are relevant to human. This would be an opportunity to bridge in between cell assays and organism level.
4/20/2016 17INTERNAL
What are barriers and limitation in using NAM in read-across?
• Difficult to make general guidance how to apply NAM evidence and how to understand what is fit for purpose.
• Cost-efficiency, how to stimulate manufacturers to use alternatives, too much expertise and resources needed.
• Almost impossible to get an overview and understanding of all NAM and how they can be applied and provide useful data.
• The adverse effect is a complex process including multiple cellular networks and dose response dependence difficult to interpret as well as relevant time points for sampling.
4/20/2016 18INTERNAL
4/20/2016 19
Conclusions from break-out session 2
How to progress together:
• Template to provide data for read-across assessment.
• Training and education.
• Develop a REACH submission based on a case study developed with NAM for example under the EU-ToxRiskproject, but with invitation to other groups of expertise to collaborate.
Break-out session 3
Case study from BASF
Read-across with metabolomics for phenoxy herbicides
Chair: Dr Tomasz Sobański, ECHA, Finland
Presenter: Dr Bennard van Ravenzwaay, BASF, Germany
Rapporteur: Dr Karel de Raat, advisor to ECHA,
The Netherlands
Topical Scientific Workshop on
New Approach Methodologies in Regulatory Science
19 – 20 April, 2016
4/20/2016 21
Question block 1: Case description
1. Are the chemical structures from the source substances and target substances sufficiently described to determine the structural (dis)similarities?
No real problems. It is clear that the substances are not the racemic mixtures of the past. However, it is not clear from the text which optical isomers are the subject of the read-across. This can easily be addressed. One participant desired a better underpinning of the choice of the substances based on chemometrics.
2. What is the regulatory purpose for which the case is prepared?
It is not clear from the document what the regulatory purpose is. The substances are herbicides with a lot of data. They have just been chosen to demonstrate the possibilities of the NAM. So in itself the case does not serve a regulatory purpose. However, the case was discussed as if it was aimed at meeting a regulatory data requirement, and not at screening or prioritization.
3. What is the property for which a prediction is attempted?
Ninety day repeated-dose oral toxicity with the rat. OECD 408
Question block 1: Case description
4. What is the hypothesis under which the prediction is attempted?
The scenario used for the RAAF assessment was either 2 for analogue approach or scenario 4 for category approach.
The actual mechanisms by which the three substances are expected to cause the same effects are established by metabolome pattern ranking. This pointed clearly to liver toxicity (PPAR alpha) and kidney toxicity. This toxicity was observed for both sources. About the same metabolome changes were found with the target substance. It was therefore concluded that this substance would cause the same effects via the same mechanisms.
The metabolome analysis combines the formulation of the hypothesis, its confirmation to an acceptable level of confidence and, in this case, a quantitative prediction.
•4/20/2016 •22
Question block 1: Case description
5. What information obtained in animal studies is provided to support the prediction?
28-day studies with the two sources and the target. However, the doses of two of them (one source and the target) were too low. ADME studies.
6. What are the weak points of the justification, which are identified by the assessment according to the RAAF?
Uncertainty about the occurrence of other effects of the target that are not reflected in its metabolome profile. The presenter emphasized during the BOS that most (nearly all?) of the effects observed in 90-day RDT studies give rise to metabolome changes and that these changes are more sensitive than the effects. However, this is not really elaborated upon in the case-study report.
•4/20/2016 •23
Question block 1: Case description
Metabolome analysis was based on 28-day studies and used for RA between 90-day studies. According to the presenter, all effects that can be seen in a 90–day study give already metabolome changes within about 14 days.
An other uncertainty was caused by the low dose-levels applied in two crucial underpinning 28-day studies.
Moreover, there was some discussion about the of effects observed. In particular effects on the testes and the adrenals, which did not fit the mechanistic explanation. However, the presenter attributed the former to reduced food intake and did not regard the latter (discoloration) as relevant/adverse.
•4/20/2016 •24
Question block 1: Case description
7. What supporting evidence is still missing or could be added to increase the confidence in the prediction?
Information on the general qualitative and quantitative sensitivity of the metabolome analysis to reflect the effects that can be observed in an oral 90-day RDT with rats.
This to build more confidence that the target will not show ‘other’effects that are not reflected in the metabolome. A validation point.
Detailed insight in the results of the/a 90-day study with 2,4 D. The absence of ‘other’effects for this substance would make the case stronger.
Combination with other NAM (bottom-up/top-down).
Etc.
•4/20/2016 •25
4/20/2016 26
Question block 2: Contribution of NAM information
8. What types of NAM information have been used in the case to increase the confidence in the prediction?
Metabolome profiling formed the core of the case. It was not an extra. Without the NAM there would not have been a viable hypothesis/mechanistic explanation.
9. Were the weak points in the prediction addressed?
Without the NAM, the case is ’empty’. It completely depends on the NAM.
4/20/2016 27
Question block 3: Detailed characteristics of the NAM information provided
10. Are the NAM methods used in the case study generally available?
The NAM depends on advanced AC analysis, statistical ‘processing’ and the availability of a database. The first two can be implemented by a well-equipped CRL. The database is not available.
More transparency is necessary to get the NAM broadly accepted for read-across purposes. The assessing scientists from regulatory agencies find it important that they can judge the validity of the methods in (much) more detail.
The current document does not provide this level of detail. This holds in particular for the database on which the NAM is used.
In this respect it is also important to mention the need of standardization.
Question block 3: Detailed characteristics of the NAM information provided
11. What scientific limitations are evident for the NAM information in the case study?
Assuming that a detailed assessment of the underlying information would not give rise to problems (see Question 10), the scientific limitations of the NAM are in THIS PARTICULAR CASE OF POSITIVE READ-ACROSS limited. The outspoken character of the results gives confidence.
Only the uncertainty of the occurrence of ‘other’ effects remains. Based on the statements of the presenter it seems that this uncertainty is not great.
•4/20/2016 •28
Question block 3: Detailed characteristics of the NAM information provided
12. What could be the barriers in using or generating the necessary NAM data?
Lack of transparency and detailed info;
Dependence of profile changes on experimental conditions and strain (see below);
The toxicological scope;
Need for standardization;
Need for validation;
Costs?????
Required technical facilities?
Required expertise (also for assessment)?
•4/20/2016 •29
Additional discussion points
The relevance of high Tanimoto scores;
Chemical read-across versus biological read-across;
Dependence of metabolome profiles on experimental conditions, in particular animal strains,
Targeted and untargeted (fingerprinting) metabolome analysis; this case is a mixture;
Acceptance of the case and scoring; large majority in favour of acceptance.
•4/20/2016 •30
4/20/2016 31
Conclusions from break-out session 3
• The case convincingly demonstrates the great potential value of metabolome analyses to support read-across;
• The analysis shows that it is highly likely that the target will cause the same effects as were observed for the sources at about the same dose level;
• There is uncertainty about possible other effects of the target that are not reflected in the metabolome;
• Implementation requires more transparency on the methods and database, additional validation studies and standardization.
Discussion on break-outsessions’ reports
Moderator:
Mr Mike RasenbergHead of Unit, Computational Assessment and Dissemination, ECHA, Finland
We are on a break and will return at 10:45
Theme 2
Screening and priority setting
Chairs:
Dr Kerry NugentNational Industrial Chemicals Notification and Assessment Scheme, Australia
Dr Jack de BruijnDirector of Risk Management, ECHA, Finland
Dr Kerry Nugent
National Industrial ChemicalsNotification and Assessment Scheme, Australia
The NICNAS IMAP Program
April, 2016
Dr Kerry Nugent – Principal Scientist
Existing Chemicals Program
NICNAS Australia
NICNAS
• National Industrial Chemicals Notification and Assessment Scheme (NICNAS)
– Commenced 1990
– Covers both human health and environment
– Manages Australian Inventory of Chemical Substances
– Premarket assessment of new chemicals
– Existing Chemicals assessed on priority basis (PEC)
• Includes ingredients in formulated cosmetics
Existing Chemicals Program Review
• Conducted 2003-2006
• Primary recommendation was to accelerate prioritisation and assessment of 38000 grandparented chemicals
• Implementation planning and program design phase
• Inventory Multitiered Assessment and Prioritisation (IMAP) program commenced June 2012
• 3000 chemicals identified for assessment
Beyond Prioritisation
• Minimum read across/QSAR hazard data objective for all chemicals
– Not “no data, no hazard”
• Integration of exposure and hazard throughout
• Preparation of a report on available information for most chemicals
• Where possible, use this information to support risk management recommendations
Exposure data (or otherwise!)
• Canada had use/volume data from DSL compilation
– No similar data for AICS
• After consultation, program designed around lack of “real” data
– Uses inferred from international information
– Mostly default volume
• Based on combination of use and volume
– See Nazaroff et al 2012 (EHP Vol 120 p1678)
Three level problem formulation
1. Is there a problem?
– Can concerns be dismissed based on readily available information?
2. Identify magnitude of problem.
– Assemble known data
– Look for straightforward resolutions based on objective knowledge
3. If needed, fully analyse problem and potential solutions.
Tier I
Tier II
Tier III
The role of the levels
• Tier I – identify chemicals not needing resourcing (de-cluttering!)
– Lack of hazard
– Lack of use
– Hazards controlled given known uses
– End result is statement of Tier I criteria met
• Tier II – Provide more information
– Identification of chemicals for which risk management can control known risk
– Propose further assessment for remainder
Tier I Assessment
Tier II Assessment
Tier III Assessment
• Chemical safety information published for community, industry and government
• Risk management recommendations supported by objective data
• Risk management recommendations for chemicals of concern
• Detailed published assessments on specific issues relating to these chemicals
• Community and industry can identify safe chemicals
• High level assessment information published
Outcomes
Tier I Matrix
Exposure Bands
• Use broadly categorised as:
– Cosmetic (100% of chemical available for exposure)
– Domestic
– Commercial
– Site Limited (0.1% available for exposure)
• Multiplied by introduction volume (known for some chemicals including HVIC) to obtain score
• Under default volumes, cosmetics = band 4 to site limited = band 1
Hazard Bands
Hazard band Hazard indicators (aligned with GHS)
Hazard band 4 CarcinogenicMutagenicReproductive/developmental toxicityEndocrine disruptionNeurotoxicity
Hazard band 3 Fatal (acute)Significant Toxicity (repeat)Respiratory sensitisationFull thickness destruction of skin tissue (limited exposure)
Hazard band 2 Toxic (acute)Harmful (repeat)Skin sensitisationFull thickness destruction of skin tissue (prolonged exposure)Eye damage
Hazard band 1 Harmful (acute)Irritant
Product Life Cycle
• In IMAP Tier I, product life cycle may be relevant
– Chemical manufacture and formulation
– Use by consumers only in dilute formulations
• May consider the relevant hazard information for the chemical as used during each life cycle stage
Risk 21 comparison
• Clearly a different matrix to Risk 21
– Does not meet the criterion of same scale on both axes
– Based on even lower data availability
• However, covers a broader range of hazards, quantitative and non-quantitative
• Uses surrogates for exposure and hazard
• Can be considered a lower tier again
Tier II
• More “classical”
• Provides information for stakeholders
• Considers whether existing risk management is appropriate
• Proposes risk management where this can be done on objective grounds
– Classification
– “Scheduling”
Objective data
• Much of the data sourced in IMAP can be considered “objective”
– Results of well-reported guideline studies
– Conclusions from well-regarded international sources
• These data can be referred to risk management agencies for consideration under their processes
Quantification
• IMAP – almost total lack of quantitative exposure data
• Can determine chemicals to be priorities to obtain data from industry…
• But can justify risk management in many cases without quantification
– “Scenarios”
What do we mean by
“risk assessment”?
What do we mean by
“risk assessment”?
Risk Assessment
Chemical
Risk
Exposure
Assessment
Hazard
Characterisation
Dose
Response
Risk Assessment
Chemical
Risk
Scenario
Identification
Hazard
Characterisation
Dose
ResponseQuantificationXX
“Risk”
Scenarios
• Almost total lack of volume and concentration data
• Uses largely estimated from international sources
• Are there circumstances where the identified hazards lead to clear risks under identified scenarios?
– Sensitising hair dyes
• Risk shown to be real but magnitude uncertain
Hazard characterisation
• Objective data
• Expert judgement using:
– Read across/grouping
– (Q)SAR (mostly mechanistic support)
– Bioelution (metals)
– Other hazard data/mechanistic
• Can accommodate future data types
– Particularly hazard characterisation
Safe Work Australia1474
Environment 73
Impact for risk management
IMAP Tier II Recommendations
At February 2016 2614 recommendations were made for 2045 unique chemicals assessed at
Tier II.
1635349
161
270
199
Tier II RecommendationsTranches 1 to 16
Safe Work Australia
Scheduling
ACCC
Tier III
Environment
Classification
Public Health
Product Safety
Tier III
Environment
Further risk management?
• Good outcomes where:
– Risk management aligned with GHS criteria
– Non-quantitative hazard eg sensitisation, genotoxic effects
– Proposal for adoption of international standard (eg SCCS)
– When seen to be proportionate
• May need further assessment and information gathering for quantification
• Question – revisiting control banding/TTC using more up to date hazard characterisation?
www.nicnas.gov.au 1800 638 528
Questions?
Dr Richard Judson
Endocrine Disruptor ScreeningProgram, United States Environmental Protection Agency, USA
Office of Research and Development
Application of computational and high-throughput in vitro screening for prioritization
Richard JudsonU.S. EPA, National Center for Computational ToxicologyOffice of Research and Development
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
ECHA Read-Across Workshop
19-20 April 2016, Helsinki
20 minutes
Office of Research and DevelopmentNational Center for Computational Toxicology
Major Points
• EDSP has a mismatch between resources needed for
Tier 1 and number of chemicals to be tested
–~10,000 chemicals in EDSP Universe
–~$1M per chemical for Tier 1, 50-100 year backlog
• Need new approach
–Prioritize chemicals
–Replace low-throughput assays with high-throughput variants
• Demonstrate new approach: Estrogen receptor
–Multiple high-throughput in vitro assays
–Demonstrate use to prioritize chemicals and replace selected
Tier 1 assays
•64
Office of Research and DevelopmentNational Center for Computational Toxicology
In Vitro Estrogen Receptor ModelCombines results from multiple in vitro assays
•65
• Use multiple assays per pathway
• Different technologies
• Different points in pathway
• No assay is perfect
• Assay Interference
• Noise
• Use model to integrate assays
• Evaluate model against reference chemicals
• Methodology being applied to other pathways
Judson et al: “Integrated Model of Chemical Perturbations of a Biological Pathway
Using 18 In Vitro High Throughput Screening Assays for the Estrogen Receptor” (submitted)
Office of Research and DevelopmentNational Center for Computational Toxicology
Immature Rat: BPA
In vivo guideline study uncertainty26% of chemicals tested multiple times in the
uterotrophic assay gave discrepant results
Kleinstreuer et al. EHP 2015
LE
L o
r M
TD
(m
g/k
g/d
ay)
Injection Oral
InactiveActive
Uterotrophic
species /
study 1
species /
study 2
Reproduce Does Not
Reproduce
Fraction
Reproduce
rat SUB rat CHR 18 2 0.90
rat CHR dog CHR 13 2 0.87
rat CHR rat SUB 18 4 0.82
rat SUB rat SUB 16 4 0.80
rat SUB dog CHR 11 4 0.73
mouse CHR rat CHR 11 4 0.73
mouse CHR rat SUB 13 7 0.65
dog CHR rat SUB 11 6 0.65
dog CHR rat CHR 13 8 0.62
rat CHR mouse CHR 11 11 0.50
mouse CHR dog CHR 6 6 0.50
rat SUB mouse CHR 13 14 0.48
dog CHR mouse CHR 6 8 0.43
mouse CHR mouse CHR 2 3 0.40
Phenotype X
Office of Research and DevelopmentNational Center for Computational Toxicology
In vitro assays also have false
positives and negatives
Much of this “noise” is reproducible
- “assay interference”
- Result of interaction of chemical
with complex biology in the assay
EDSP chemical universe is structurally
diverse
-Solvents
-Surfactants
-Intentionally cytotoxic compounds
-Metals
-Inorganics
-Pesticides
-Drugs
Assays cluster by technology,
suggesting technology-specific
non-ER bioactivity
Judson et al: ToxSci (2015)
Chem
icals
Assays
Office of Research and DevelopmentNational Center for Computational Toxicology
Assay-to-assay variation
•68
Agonist
Antagonist
All appropriate
assays are active
but efficacy and
potency vary
“Noise” or real
variation in biology
between cell types?
Judson et al: ToxSci (2015)
Assay Data Integrated Model
Office of Research and DevelopmentNational Center for Computational Toxicology
In Vitro Reference
Chemical Performance
Office of Research and DevelopmentNational Center for Computational Toxicology
Uterotrophic Database
98 Chemicals
442 GL uterotrophic bioassays
Literature Searches:
1800 Chemicals
Data Review:
700 Papers, 42 Descriptors, x2
6 Minimum
Criteria
High-Level
Filter
In Vivo ER Reference Chemicals
30 Active, 13 Inactive
Identifying Uterotrophic Reference
Chemicals from the Literature
Selection
Criteria
“Guideline-Like”
(GL)
Kleinstreuer et al: “A Curated Database of Rodent Uterotrophic Bioactivity” (submitted)
Office of Research and DevelopmentNational Center for Computational Toxicology
Model predicts in vivo uterotrophic assay as well
as uterotrophic predicts uterotrophic
Rank Order (ER Agonist AUC)
Kaempferol
Active
InactiveUterotrophic
D4
Restrict to chemicals with consistent
results from the literature
Browne et al. ES&T (2015)
ER
Ag
on
ist
AU
C
True Positive 29
True Negative 50
False Positive 1
False Negative 1
Accuracy 0.97
Sensitivity 0.97
Specificity 0.98
Office of Research and DevelopmentNational Center for Computational Toxicology
Explicitly Add Uncertainty to In Vitro Assay Data
•72
Watt et al. (in prep)
Quantify
uncertainty
Consensus
model
Office of Research and DevelopmentNational Center for Computational Toxicology
CERAPP: using QSAR for further prioritization
• Collaborative Estrogen Receptor Activity Prediction Project
• Goals:
–Use ToxCast ER score (or other data) to build many QSAR models
–Use consensus of models to prioritize chemicals for further testing
• Assumptions
–ToxCast chemicals cover enough of chemical space to be a good
“global” training set
–Consensus of many models will be better than any one individually
• Process
–Curate chemical structures
–Curate literature data set
–Build many models
–Build consensus model
–Evaluate models and consensus•73
Mansouri et al: “CERAPP: Collaborative Estrogen Receptor Activity Prediction Project” EHP (2016)
Office of Research and DevelopmentNational Center for Computational Toxicology
Total Database
Binders: 3961
Agonists: 2494
Antagonists: 2793
CERAPP Consensus evaluation
Key point: As greater consistency
is required from literature sources,
QSAR consensus model
performance improves
Office of Research and DevelopmentNational Center for Computational Toxicology
CERAPP Summary
• EDSP Universe (10K)
• Chemicals with known use (40K) (CPCat & ACToR)
• Canadian Domestic Substances List (DSL) (23K)
• EPA DSSTox – structures of EPA/FDA interest (15K)
• ToxCast and Tox21 (In vitro ER data) (8K)
~32K unique structures
5-10% predicted to be ER-active
Prioritize for further testing
75
Office of Research and DevelopmentNational Center for Computational Toxicology
ER Phenol Read-Across Model
Filtering 1 (Log Pkow & MV) Filtering 2 (No. of Literature Sources >= 3)
Pradeep et al. (in prep)
Accuracy increases as
1. Better data is used in the evaluation
2. Neighbors are closer (structure and physchem)
Office of Research and DevelopmentNational Center for Computational Toxicology
Moving Towards Regulatory Acceptance
From FIFRA SAP, December 2014
• Can the ER Model be used for prioritization?
– “… the ER AUC appears to be an appropriate tool for chemical prioritization for …
the EDSP universe compounds.”
• Can the ER model substitute for the Tier 1 ER in vitro and uterotrophic
assays?
– “… replacement of the Tier 1 in vitro ER endpoints …with the ER AUC model will
likely be a more effective and sensitive measure for the occurrence of estrogenic
activity …”
– “… the Panel did not recommend that the uterotrophic assay be substituted by
the AUC model at this time. The Panel suggested that the EPA considers: 1)
conducting limited uterotrophic and other Tier 1 in vivo assay testing, using the original
Tier 1 Guidelines (and/or through literature curation)”
• Based on follow-up presented here (FR notice, June 18 2015) …
– “EPA concludes that ER Model data are sufficient to satisfy the Tier 1 ER
binding, ERTA and uterotrophic assay requirements.”
•77
Office of Research and DevelopmentNational Center for Computational Toxicology
Data Transparency: EDSP21 Dashboard
• Goal: To make EDSP21 data easily available to all
stakeholders
–Assay-by-assays concentration-response plots
–Model scores – AUC agonist and antagonist
–ER QSAR calls
–Other relevant data
• https://actor.epa.gov/edsp21
Office of Research and DevelopmentNational Center for Computational Toxicology
Summary
• EDSP is in need of new approach to handle large
testing universe
–Reduce cost, speed throughput
• Estrogen Receptor Model is first example of this
–54 chemicals in low-throughput Tier 1 assays
–1800 chemicals tested and published in high-throughput
–1000 more in queue – 2016 planned release
• Next steps
–Androgen receptor (1800 chemicals tested, modeling and
validation in progress)
–Steroidogenesis (1000 chemicals with preliminary data)
–Thyroid – assay development and testing underway for several
targets (THR, TPO, deiodinases, ...)
•79
Office of Research and DevelopmentNational Center for Computational Toxicology
AcknowledgementsNCCT Staff Scientists
Rusty Thomas
Kevin Crofton
Keith Houck
Ann Richard
Richard Judson
Tom Knudsen
Matt Martin
Grace Patlewicz
Woody Setzer
John Wambaugh
Tony Williams
Steve Simmons
Chris Grulke
Jim Rabinowitz
NCCT
Nancy Baker
Jeff Edwards
Dayne Filer
Parth Kothiya
Doris Smith
Jamey Vail
Sean Watford
Indira Thillainadarajah
NIH/NCATS
Menghang Xia
Ruili Huang
Anton Simeonov
NTP
Warren Casey
Nicole Kleinstreuer
Mike Devito
Dan Zang
Kamel Mansouri
Nicole Kleinstreuer
Eric Watt
Prachi Pradeep
Patience Browne
NCCT Postdocs
Todor Antonijevic
Audrey Bone
Kristin Connors
Danica DeGroot
Jeremy Fitzpatrick
Jason Harris
Dustin Kapraun
Agnes Karmaus
Max Leung
Kamel Mansouri
LyLy Pham
Prachi Pradeep
Caroline Ring
Eric Watt
Questions?