Toxicogenomics; toepasbaarheid van deze methodieken voor de praktijk Danyel Jennen Dept of Toxicogenomics Maastricht University, the Netherlands
Toxicogenomics Toxicologie
- gevaar identificatie - risico karakterizering - mechanismen - dosis-respons effecten - “modifier” effecten - ziekte resultaat - blootstellingsbeoordeling
Systeem Biologie
-
Genomics Genetics Epigenomics Transcriptomics Proteomics Metabolomics Bioinformatica
Toxicogenomics projecten
ASAT2 DECO / DECO2 Cefic LRI-AIMT3 / AIMT4
In vitro predictie modellen
Genotoxiciteits en carcinogeniciteits beoordeling van chemicaliën Chemicalie
Gene mutations Chromosomal mutations DNA schade
Translocaties Breuken Micronuclei Sister chromatid exchanges
Genotoxiciteits tests in vitro en in vivo
Kanker
Knaagdier Carcinogeniciteits tests
Wanneer toxicogenomics toepassen voor Genotoxiciteit / Carcinogeniteit Chemicalie In vitro GTX tests: Gen mutaties (bacteria + mammalian cells) Chromosomale mutaties In vivo GTX tests: Gen mutaties Chromosomale mutaties DNA repair Knaagdieren (muis + rat) Carcinogeniteits tests
Voor GTX en Carcinogeniteit Reductie alle diertests
Voor in vivo GTX en Carcinogeniteit Reductie alle diertests
Voor Carcinogeniteit Reductie Carc. tests
Transcriptomics-gebaseerde assay het basis principe Genotoxische stoffen (GTX)
Non-genotoxische stoffen (NGTX)
Onbekende stof
A B Negatief GTX
Transcriptomics-gebaseerde assay tijdslijn
Gepatenteerd:
EPO Patent EP2525223 A1 WO 2012/156526
9
Algemene opzet Celkweek
• HepG2
Blootstelling
Gen expressie
Predictie
• Affymetrix • Gen selectie • Stoffen: whole Genome – T-tests p<0.01 – GTX / NGTX – Leave-1-out – Carc / NCarc – Training en validatie • Predictie – PAM software • 2-3 bloodstellingstijden: 12, 24, 48u – Leave-1-out • 1 dosis: MTT-IC20 or 2 mM • Classificatie als • 3 onafhankelijke experimenten 2/3 of 3/3
Genotoxiciteits predictie methodes: stratificatie van stoffen gebaseerd op in vitro GTX tests
Method 1
Method 3
Method 2
100%
100%
100%
80%
80%
80%
60%
60%
60%
40%
40%
40%
20%
20%
20%
0%
0% 12h Accuracy
24h Sensitivity
48h Specificty
0% 12h Accuracy
24h Sensitivity
48h Specificity
12h Accuracy
24h Sensitivity
48h Specificity
Genotoxiciteits predictie methodes: stratificatie van stoffen gebaseerd op in vitro GTX tests
Method 1
Validation 24h
100%
100%
100%
80%
80%
80%
60%
100%
40% 20%
40%
80%
80%
60%
40%
60%
40%
20%
40%
20%
0%
20%12h 0%
Accuracy
24h Sensitivity
Accuracy
Specificty
Method 1 Accuracy
0% 12h
48h
Method 2 Sensitivity
Method 3 Specificity
Method 3
100%
60%
60%
0%
Validation 48h
Method 2
24h 20%
48h
12h
Sensitivity 0%Specificity
Accuracy
Method 1
24h Sensitivity
Method 2 Accuracy
Sensitivity
48h Specificity
Method 3 Specificity
Table 3: Comparison of the performance for predicting in vivo genotoxicity of the transcriptomicsbased assay upon 24h of exposure and Ames stratification of chemicals with conventional in vitro genotoxicity assays and combinations thereof Ames + MLAb
Ames + GEb
MN/C Ac
Ames + MN/CAc
Ames + GEc
60% 94%
60% 94%
91% 100%
63% 96%
62% 96%
9%
6%
6%
0%
4%
77%
87%
42%
42%
97%
23%
13%
58%
58%
3%
Ames + GEa
MLA
77% 78%
89% 91%
22%
Ames a
Accuracy Sensitivity False negative rate Specificity False Positive rate
b
Ames + MLA/MN/CA d
Ames + GEa
88% 91%
68% 96%
89% 91%
4%
9%
4%
9%
46%
40%
86%
51%
87%
54%
60%
14%
49%
13%
MLA: Mouse Lympoma Assay, GE: Gene expression, MN: Micronuclei Assay, CA: Chromosomal Aberrations a: based on 62 compounds with available Ames results; b: based on 47 compounds with available MLA results; c: based on 60 compounds with available MN or CA (or both results); d: based on 62 compounds with data in at least one of the four in vitro assays.
in vitro – in vivo vergelijking
Connectivity mapping
Lamb et al. Science. 2006 Sep 29;313(5795):1929-35
Connectivity mapping
Lamb et al. Science. 2006 Sep 29;313(5795):1929-35
Cmap method
19070
19070
+1
3 2 1
1 -2 -3
………......
-1
……………
19069
REFERENCE
Reassign signs
………………………
GENESET
C
Absolute ranking
-19069
Cmap method
19070
19070
+1
3 2 1
1 -2 -3
………......
-1
……………
19069
REFERENCE
Reassign signs
………………………
GENESET
C
Absolute ranking
-19069
Reference data : TG-Gates
n= 3 + 23
n=91
n=15
n=16
HHC &
Liver
RHC Group1 (A and B)
HC RC
NOT Liver
Group2
NC
NO target
Group3
NK
NO target
Group4
Query signature : genesets 112 genesets for HCC
9 Databases (LOMA, CTD)
63 Literature
23 Metacore
17 Pathways
31 genesets selected
5 Databases
24 Literature
2 Pathways
Query signature : in vivo
Query signature : genesets
Data integratie & read across
iClusterPlus
DECO / DECO2
• Predictie van kanker subtypes • Integratie discrete en continue variabelen
http://www.mskcc.org/research/epidemiology-biostatistics/biostatistics/iclusterplus R package: iClusterPlus
Latente variabelen of Verborgen motieven
iClusterPlus op structureel vergelijkbare stoffen van DrugMatrix Tanimoto score > 0.8
1=betamethasone 2=dexamethasone 3=hydrocortisone 4=17-methyltestosterone 5=ethisterone 6=norethindrone acetate 7=progesterone 8=cortisone 9=testosterone
Clinical chemistry
Transcriptomics
iClusterPlus op DrugMatrix data, Transcriptomics (144 features) & Clinisch chemisch (613 features)
data infrastructure for chemical safety www.dixa-fp7.eu/home
data infrastructure for chemical safety www.dixa-fp7.eu/home
data infrastructure for chemical safety www.dixa-fp7.eu/home
diXa can serve as a framework to ensure that future Toxicogenomics based research projects produce clear, documented useable codes in their data management/open accessibility of the generated research data. John Quackenbush, Professor of Computational Biology and Bioinformatics:”… We need a) methods to compare measurements across sources and rapidly identify salient features, b) methods that can combine data from various sources where there are hidden correlations in the data c) leverage the volume and velocity of the data to provide opportunities for validation of findings and d) move beyond correlations in studying relationships in data …” and diXa clearly contributes to these big data challenges.
data infrastructure for chemical safety www.dixa-fp7.eu/home
Richard Currie, Syngenta: “… diXa organises and integrates toxicogenomics data in such a way that it permits us to use this data infrastructure in safety assessment …” Garry Miller, Editor-in-Chief Toxicological Sciences: “… diXa created a data warehouse framework, a collection of European Toxicogenomics experiments with cross-links to chemical and molecular medicine databases which is, to my knowledge, the best known example in data management in the field of Toxicogenomics …” Joop de Knecht, Environment Directorate OECD: “…. It is a lot easier to predict that a chemical will cause an effect than to predict that it is save, a project like diXa contributes to integrate differences in responses to chemicals in risk assessment …”
Het ontstaan van diXa’s data infrastructuur * door integratie van TGX data van FP6/FP7 projecten * door koppeling met chemische/mol. medicijn data bases * dus mogelijk maken van cross-platform, cross-studie integrations
diXa’s Data Warehouse www.dixa-fp7.eu/home
Acknowledgement
