McKim Workshop on Strategic Approaches for Reducing Data Redundancy in Cancer Assessment
Jay R. Niemelä Technical University of Denmark
National Food InstituteDivision of Toxicology and Risk Assessment
e-mail: [email protected]
In silico methods for predicting chromosomal endpoints for carcinogens
2 DTU Food, Technical University of Denmark
Eva Bay WedebyeGunde Egeskov Jensen
Marianne DybdahlNikolai NikolovSvava Jonsdottir
Tine Ringsted
3 DTU Food, Technical University of Denmark
Data set: EINECS 49,292 discrete organics• European Inventory of Existing Chemical Substances
• Very similar to U.S TSCA inventory and expected to contain most REACH chemicals.
4 DTU Food, Technical University of Denmark
Objective• 1. To define a large set of carcinogens and non-carcinogens
• 2. Analyse these chemicals for genotoxic potential in a set of in vitro models
• 3. Further assess performance in in vivo models.
5 DTU Food, Technical University of Denmark
Pure In Silico
Any relation to test data is incidental
6 DTU Food, Technical University of Denmark
Method
Global (Q)SARs
in between
Local (Q)SARsClosely related structuresAccurate predictions for a small number of chemicals
Fragment rule-basedFastHigh throughputDiverse
7 DTU Food, Technical University of Denmark
Model Platform: MULTICASE
• Cancer models
• MULTICASE FDA proprietary, male and female mouse and rat• MULTICASE Ashby fragments
8 DTU Food, Technical University of Denmark
Gentotoxicity models. Developed in-house. QMRF’s and training sets availableIn Vitro• HGPRT forward mutation in CHO cell• Mutations in mouse lymphoma • Chromosomal aberration CHL • Reverse mutation test, Ames• SHE cell transformation
In Vivo• Drosophila melanogaster Sex-Linked Recessive Lethal • Mutations in mouse micronucleus • Dominant lethal mutations in rodent • Sister chromatid exchange in mouse bone marrow• COMET assay in mouse
9 DTU Food, Technical University of Denmark
Domaine• Only predicitons with no fragment- or statistical warnings were used.
• For positive cancer predictions, ICSAS criteria, meaning that at least two were positive (trans-gender or trans-species)
• To be considerd a non-carcinogen, chemicals had to be predicted negative in all four models (MM, FM, MR, FR)
10 DTU Food, Technical University of Denmark
Activity distribution
6177
27362
15753
0
5000
10000
15000
20000
25000
30000
Positive Unpredicted Negative
11 DTU Food, Technical University of Denmark
Clustering actives
12 DTU Food, Technical University of Denmark
Structures
13 DTU Food, Technical University of Denmark
Activity distribution with Ashby positives removed
4037
27362
15753
2140
0
5000
10000
15000
20000
25000
30000
Positive Unpredicted Negative
14 DTU Food, Technical University of Denmark
In vitro results for Ashby negative carcinogens
Ames CA ML HGPRT UDS SHE
Ames 934 159 504 293 91 345
CA 516 189 101 45 103
ML 1167 395 116 472
HGPRT 559 80 288
UDS 259 87
SHE 768
15 DTU Food, Technical University of Denmark
General estimates and in vitro predictions (4037)
Ames test 934 (21.1%)
Chromosomal aberrations 516 (12.8%)
Mouse lymphoma 1167 (28.9%)
HGPRT 559 (13.8%)
Unscheduled DNA synthesis 259 (6.4%)
Cell transformation (SHE) 768 (19.0%)
16 DTU Food, Technical University of Denmark
In vitro mutagens
Predicted positive in Ames test, Mouse lymphoma, or Chromosomal aberrations CHL
Non-mutagens 2184
Mutagens 1853
Non-mutagens Mutagens
17 DTU Food, Technical University of Denmark
Distribution of in vivo positives (1853)
1853 Genotoxic carcinogens
15753 Non-carcinogens
Mouse micronucleus 231 1640
Sister chromatid exchange 800 2671
Comet assay 288 2330
Drosophila sex-linked recessive lethal 77 550
Rodent dominant lethal 102 741
18 DTU Food, Technical University of Denmark
Distribution of in vivo positives by percent
Genotoxic carcinogens, %
Non-carcinogens, %
Mouse micronucleus 12.5 10.4
Sister chromatid exchange 43.2 17.0
Comet assay 15.5 14.8
Drosophila sex-linked recessive lethal 4.2 3.5
Rodent dominant lethal 5.5 4.7
19 DTU Food, Technical University of Denmark
In vivo models as predictors of genotoxic carcinogenicity AM CA ML (1853)
Model utility (TP - FP) shown by red bars
0 10 20 30 40 50
SCE
MM
DL
COMET
SLRL
FP TP TP - FP
20 DTU Food, Technical University of Denmark
In vivo models as predictors of carcinogenicity - Cell transformation SHE (768)
Model utility (TP - FP) shown by red bars
-10 0 10 20 30 40 50 60
SCE
COMET
MM
DL
SLRL
FP TP TP - FP
21 DTU Food, Technical University of Denmark
Cluster of SHE/SCE positives
22 DTU Food, Technical University of Denmark
Activity distribution with Ashby negatives removed
4037
27362
15753
2140
0
5000
10000
15000
20000
25000
30000
Positive Unpredicted Negative
23 DTU Food, Technical University of Denmark
In vitro results for Ashby positive carcinogens
Ames CA ML HGPRT UDS SHE
Ames 918 472 498 336 160 349
CA 944 434 319 110 343
ML 982 412 128 383
HGPRT 496 86 253
UDS 230 80
SHE 560
24 DTU Food, Technical University of Denmark
General estimates and in vitro predictions (2140)
Ames test 918 (42.9%)
Chromosomal aberrations 944 (44.1%)
Mouse lymphoma 982 (45.9%)
HGPRT 496 (23.2%)
Unscheduled DNA synthesis 230 (10.7%)
Cell transformation (SHE) 560 (26.2%)
25 DTU Food, Technical University of Denmark
In vitro mutagens from Ashby positives
Predicted positive in Ames test, Mouse lymphoma, or Chromosomal aberrations CHL
Non-mutagens 437
Mutagens 1703
Non-mutagens Mutagens
26 DTU Food, Technical University of Denmark
Distribution of in vivo positives (1703)
1703 Genotoxic carcinogens
15753 Non-carcinogens
Mouse micronucleus 272 1640
Sister chromatid exchange 649 2671
Comet assay 458 2330
Drosophila sex-linked recessive lethal 194 550
Rodent dominant lethal 159 741
27 DTU Food, Technical University of Denmark
Distribution of in vivo positives by percent
Genotoxic carcinogens, %
Non-carcinogens, %
Mouse micronucleus 16 10.4
Sister chromatid exchange 38.1 17.0
Comet assay 26.9 14.8
Drosophila sex-linked recessive lethal 11.4 3.5
Rodent dominant lethal 9.3 4.7
28 DTU Food, Technical University of Denmark
In vivo models as predictors of genotoxic carcinogenicity AM CA ML (1703)
Model utility (TP - FP) shown by red bars
0 10 20 30 40 50
SCE
COMET
SLRL
MM
DL
FP TP TP - FP
29 DTU Food, Technical University of Denmark
Conclusions:
”Fragment” or ”Rule-Based ” systems provide extremely valuable information, particularly for genotoxic carcinogens
In Silico methods could help scientists looking for new fragments or rules
Current regulatory use of in vivo tests may need to be modified if they are going to replace carcinogenicity bioassays