Health Canada experiences with early Health Canada experiences with early identification of potential carcinogensidentification of potential carcinogens- An Existing Substances Perspective- An Existing Substances Perspective
SSunil Kulkarniunil KulkarniHazard Methodology Division, Hazard Methodology Division,
Existing Substances Risk Assessment BureauExisting Substances Risk Assessment Bureau
Health Canada, Ottawa, ONHealth Canada, Ottawa, ON
OutlineOutline
• Brief introduction• DSL - Categorization – Tools/Approaches• Chemicals Management Plan – Phase I & II• Remaining priorities• (Q)SAR tools we use • Challenges of (Q)SAR models & modelable endpoints• (Q)SAR results/analyses
Existing Substances under CEPA 1999Existing Substances under CEPA 1999
• Approximately 23,000 substances (e.g., industrial chemicals) on the Domestic Substances List (DSL)
• Includes substances used for commercial manufacturing or manufactured or imported in Canada at >100 kg/year between Jan 1, 1984 and Dec 31, 1986
Categorization Categorization
• Identify substances on the basis of exposure or hazard to consider further for screening assessment and to determine if they pose “harm to human health” or not
• A variety of tools including those based on (Q)SAR approaches were applied
~3200 ~3200 remaining remaining prioritiespriorities
Categorization
23,00023,000DSL DSL chemicalchemicals s
4,300 priorities
Chemicals Management Plan
Chemicals Management Plan (CMP)Chemicals Management Plan (CMP)• To assess and manage the risks associated with 4300 legacy substances identified through categorization by 2020
• 4300 substances were prioritized into high (~500), medium (~3200) and low concern substances (~550)
• CMP brings all existing federal programs together into a single strategy to ensure that chemicals are managed appropriately to prevent harm to Canadians and their environment
•It is science-based and specifically designed to protect human health and the environment through four major areas of action:
• Taking action on chemical substances of high concern• Taking action on specific industry sectors• Investing in research and biomonitoring• Improving the information base for decision-making through
mandatory submission of use and volume information
DSL Categorization Commercial (Q)SAR models; basis for decision making (prioritization)
2000-06
Commercial and some public domain (Q)SAR models, Metabolism, Analogue identification, Read-across; basis for decision making but mainly supportive evidence
Ministerial Challenge PhaseCMP (high priorities)
2006-11
2011- CMP II(includes data poor substances)
Commercial and public domain (Q)SAR models, Analogue identification, Chemical categories, Read-across, Metabolism, in-house models/tools
Historical use of (Q)SAR applicationsHistorical use of (Q)SAR applications
Included in Rapid Screening: 545
Addressed through PBiT SNAcs: 145
Being addressed in Petroleum Sector Stream Approach: 164
Addressed in the Challenge: 200
Remaining priorities to be addressed by 2020: 3200
Universe of chemicals in work planUniverse of chemicals in work plan4300 existing chemical substances to be addressed by 2020:
~1500 to be addressed by 2016 through the groupings initiative, rapid screening and other approaches
Remaining Priorities - ScopeRemaining Priorities - Scope
(Q)SAR tools are generally only (Q)SAR tools are generally only applicable to discrete organics!applicable to discrete organics!
Remaining Priorities – Data availabilityRemaining Priorities – Data availability
Are there enough data-rich analogues?
(Q)SAR opportunities?
58%
4%15%
23%
ApproachApproach
Human health risk assessmentHuman health risk assessment• Chemical’s inherent toxicity & potential human exposure
• Assess a range of endpoints including genotoxicity, carcinogenicity, developmental toxicity, reproductive toxicity & skin sensitization
• (Q)SAR approaches, including analogue/chemical category read across are used to support our assessments (line of evidence)
• Apply weight of evidence and precaution in our decision-making
Hierarchical consideration of sources Hierarchical consideration of sources of informationof information
Chemical
Predictive tools for hazard assessmentPredictive tools for hazard assessment
Commercial
• Casetox• Topkat• Derek • Model Applier• Oasis Times
Non-commercial
• OECD QSAR Toolbox• Toxtree • OncoLogic• Caesar (Vega)• lazar
Supporting tools• Leadscope Hosted - chemical data miner• Pipeline Pilot – cheminformatics and workflow builder
Identifying toxic potentialIdentifying toxic potential
Relevance to humans
Relevance to humans
Essential to have a balanced judgement of the totality of available evidence
Expert systems
Chemical of interest
In vitrodata
In vivo mammalian
data
QSAR models
Analogue/Chemical
category read across
Toxic potential
Sufficient information
Insufficient information
Sufficient information
Hazard assessment
Reliability of estimationsReliability of estimations• Minimizing uncertainties and maximizing confidence in
predictions considering multiple factors:
- OECD QSAR Validation principles - accuracy of input - quality of underlying biological data - multiple models based on different predictive paradigms or
methodologies- mechanistic understanding- inputs from in vitro/in vivo tests (if available)
• Professional judgement of expert(s)
(Q)SAR tools/approaches to identify (Q)SAR tools/approaches to identify potential potential genotoxicgenotoxic carcinogens carcinogens
• QSAR Toolbox profiler flags- DNA/Protein binding, Benigni-Bossa, OncoLogic
• Metabolic simulators (Toolbox/TIMES) + DNA/Protein binding/Benigni-Bossa flags
• Combination of (Q)SAR models for genotoxicity & carcinogenicity (Casetox, Model Applier, Derek, Times, Toxtree, Caesar, Topkat)
• Genotox - Salmonella (Ames) models for different strains, Chrom ab, Micronuclei Ind, Mouse Lymphoma mut with metabolic activation
• Carcinogenicity – Male & female rats, mice, rodent
(Q)SAR tools/approaches to identify (Q)SAR tools/approaches to identify potential potential non-genotoxicnon-genotoxic carcinogens carcinogens
• Flags from QSAR Toolbox profilers – Benigni-Bossa
flags
• QSAR models based on in vitro Cell Transformation
assays such as Syrian Hamster Embryo, BALB/c-3T3,
C3H10T1/2
• Expert rule based systems Derek and Toxtree
In vitro CTA
In vivomammalian
(Q)SAR/Read across
In vivo/in vitro
Genotox
Expert rules/knowledge
Male rat
Female rat
Male mice
Female mice
ChromAber
Micronuclei Ind
Mouse Lymphoma
Salmonella Ames
Drosophila
Unsch DNA Syn
SisterChrExc
SHE
BALB/c-3T3
C3H10T1/2
Non-genotoxic
DNA binding
Protein binding
Metabolism
Genotoxic
Holds potential to Holds potential to form part of hazard form part of hazard identification identification strategystrategy
Helpful to have a better Helpful to have a better understanding of Cell understanding of Cell
Transformation information in Transformation information in mechanistic interpretation of mechanistic interpretation of
(non-genotoxic) carcinogenicity(non-genotoxic) carcinogenicity
Domain of Domain of most most
(Q)SAR (Q)SAR modelsmodels
Few or no Few or no robust robust (Q)SAR (Q)SAR modelsmodels
Ashb
y (1
992)
, Pre
dicti
on o
f non
-gen
otox
ic c
arci
noge
nesi
s. T
oxic
olog
y Le
tter
s, 6
4/65
, 605
-612
.
Few or no (Q)SAR modelsFew or no (Q)SAR models
Basis of non-empirical approachesBasis of non-empirical approachesPhysChemBio activity Function of Ability to model/
Use in decision-making
Simple Molecular structure Good
Complex
Molecular structureMechanism MetabolismMulti-step
Challenging (uncertainty ↑)
Complex BA not easily translated/explainable in terms of simple molecular structure/fragments to enable building a robust QSAR
For instance, a QSAR model for carcinogenicity only predicts Yes/No without any information about its mechanism
Availability of data rich analogues is essential for read-across approaches
(Q)SAR analysis(Q)SAR analysis
Performance of some (Q)SAR modelsPerformance of some (Q)SAR models• A set of chemicals with in vitro and in vivo data on genotoxicity
and carcinogenicity was chosen• Predictions were obtained for different human health relevant
endpoints by running these through a variety of (Q)SAR models• Performance of models to discriminate carcinogenic and non-
carcinogenic chemicals was evaluated by analysing the results • Structural analysis of chemicals incorrectly classified by all
models revealed a diverse group of chemicals with few trends (we are working on that)
• Failure of models/expert systems to flag them as “Out of domain”
Prediction results/analysisPrediction results/analysis
Dataset of approx. 100 chemicals:
Ames PN ratio=55:46 Carc PN ratio: 49:52. 23 are positive in both Carc and Ames20 are negative in both; 32 are only Ames positive26 are Carc positive but Ames negative (non-Gtx Carc?)
Model a1 a2 b1 b2 c1 c2 d SHE-NgC SHE-Carc
TP 41 35 21 21 16 12 16 11 27
TN 32 46 32 35 30 13 8 9 17
FP 18 5 6 4 6 2 11 9 17
FN 5 12 14 16 16 2 2 4 7
total 96 98 73 76 68 29 37 33 68
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
True
pos
itive
rate
False positive rate
Performance of QSAR models to discriminate Performance of QSAR models to discriminate carcinogenic/non-carcinogenic chemicals (n=100)carcinogenic/non-carcinogenic chemicals (n=100)
Models Casetox 2.4Model Applier 1.4Topkat 6.2Toxtree 2.5SHE=Syrian Hamster Embryo modelNgC=Non-genotoxic carcinogenicity
a1 (96)
a2 (98)
b1 (73)
c1 (68)b2 (76)
c2 (29)
SHE carc(68)
d (37)
Performance of Performance of in vitroin vitro Cell Transformation QSAR Cell Transformation QSAR models to discriminate carcinogenic/non-models to discriminate carcinogenic/non-
carcinogenic chemicals (n=130)carcinogenic chemicals (n=130)
LegendCTA=Cell Transformation assay based modelSHE=Syrian Hamster EmbryoBALB/c 3T3C3H 10T1/2
BALBc (115)
C3H10T1 (50)
SHE (96)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
TPR
FPR
CTA models exhibit potential but there is scope for improvement
Performance of some (Q)SAR models to Performance of some (Q)SAR models to identify non-genotoxic carcinogensidentify non-genotoxic carcinogens
Current cancer models aren’t designed to inform about genotoxic or non-genotoxic events in the carcinogenesis process
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
TPR
FPR
SHE(31)
a1(43)
a2(44)
c2(10)
b2(42)
c1(33)
b1(41)
d1(6) e(20)
d2(46)
Data analysisData analysis
Comparative ability of Ames & SHE tests to Comparative ability of Ames & SHE tests to discriminate carcinogens/non-carcinogensdiscriminate carcinogens/non-carcinogens
SHE (150)
SHE+Ames (70)
Ames (700)
0.00
0.20
0.40
0.60
0.80
1.00
0.00 0.20 0.40 0.60 0.80 1.00
TPR
FPR
MN (190)
CA (300)
MLm (220)
SHE (55)
Performance of genotoxicity and CT tests to Performance of genotoxicity and CT tests to discriminate (Ames -) carcinogens/non-carcinogensdiscriminate (Ames -) carcinogens/non-carcinogens
LegendSHE=Syrian Hamster EmbryoMLm=Mouse Lymphoma mutationCA=Chromosomal AberrationMN=Micronuclei induction
Performance of genotoxicity and CT tests to Performance of genotoxicity and CT tests to discriminate (Ames +) carcinogens/non-discriminate (Ames +) carcinogens/non-
carcinogenscarcinogens
0.00
0.20
0.40
0.60
0.80
1.00
0.00 0.20 0.40 0.60 0.80 1.00
TPR
FPR
SHE (60)
MLm(155)
CA (245)
MN (110)
Ability of reprotoxicity data to Ability of reprotoxicity data to discriminate carc/non-carc chemicalsdiscriminate carc/non-carc chemicals
FMR (27)
FRR (107)
FRodR (118)
MMR (29)
MRR (72)
MRodR (83)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
TPR
FPR
LegendFRR=female rat reproductiveFRodR=female rodent reproMMR=male mice reproFMR=female mice reproMRodR=male rodent reproMRR=male rat repro
Scope for improvementScope for improvement
Finally………..Finally………..
fpr
tpr
Examples from CMP I where (Q)SAR Examples from CMP I where (Q)SAR or analogue-read across approaches or analogue-read across approaches
were used as supporting were used as supporting informationinformation
n-butyl glycidyl ether(CAS 2426-08-6 )
N
N
S
N
N
N Cl
Cl
CH3
O
N+O–
N
N
NN
H2
CN
CH3
CH3
S
MAPBAP acetate(CAS 72102-55-7)
DAPEP (CAS 25176-89-0 )
Disperse Red 179(CAS 16586-42-8)
http://www.chemicalsubstanceschimiques.gc.ca/challenge-defi/index-eng.php