Case study: integration into risk
assessment of human omics data
from in vitro studies
AIM
• Integration of human in vitro omics data with information extracted from
adverse outcome pathways (AOPs) in order to identify areas of concern and
support an evidence-driven risk assessment
• Integration of different modelling tools
• Use only:
• Open source data
• Human data
• In vitro data
• Compound: piperonyl butoxide (PBO) - CAS Number: 51-03-6
• Insecticide synergist, is also included in cosmetic products for skin protection (CosIng)
• Classified as a non-genotoxic carcinogen
• A PBO metabolite binds to Cytochrome P450 enzymes, thus reducing the ability of
the enzymes in breaking down accompanying pesticides
• Induced increase of alkylation of macromolecules
23 September 2016
2
DATA
Name
Data Infrastructure for Chemical Safety
ArrayExpress Archive of Functional
Genomics Data
EPA Toxicity ForeCaster (ToxCast™) Data
EPA iCSS ToxCast Dashboard
Abbreviation
diXa
ArrayExpress
ToxCast
iCSS
Use
Transcriptomics data on test compound PBO
http://wwwdev.ebi.ac.uk/fg/dixa/group/DIXA-002
Transcriptomics data on reference compounds MTX and
VPA
Download high-throughput toxicity data set
Explore ToxCast data
23 September 2016
3
TOOLS
Name
EPA Aggregated Computational Toxicology
Resource
Integrated analysis of Cross-platform
MicroArray and Pathway data
Abbreviation
ACToR
InCroMap
Comparative Toxicogenomics Database
CTD
Connectivity Map
cmap
Mode of Action by NeTwoRk Analysis
Adverse Outcome Pathway Knowledge
Base
Use
Explore toxicity information
Kyoto Encyclopedia of Genes and Genomes (KEGG)
pathway analysis
Pathway-genes-diseases-chemicals associations analysis
Mantra 2.0
AOP-KB
Verification of adverse effects versus specific KE
23 September 2016
4
METHODOLOGY
23 September 2016
5
STEPS
1.
Transcriptomics data from three human hepatocytes models treated
with different concentrations of PBO for 24h or 72h was used for
KEGG pathway analysis in InCroMap software. (done)
2.
The most relevant molecular pathways which showed to be
influenced by the treatment with PBO were selected and further
analyzed for genes, diseases and chemicals associations in CTD.
(done)
3.
Connectivity map (next step)
4.
Mode of Action by NeTwoRk Analysis (MANTRA)
5.
Verification of adverse effects versus specific AOP key events (KI)
and molecular initiating events (MIE), using the information from
the AOP-Knowledge Base.
23 September 2016
6
STEPS
Connectivity Map
a. Starting point – identify appropriate transcriptomic data of PBO
•
•
PBO tested in two different cell types (HepaRG and HepG2), two different
concentrations and two time points (24h and 72h).
Data: QC, normalization, transformation and analyses of transcriptomics data
b. Find similar compounds to PBO profile, by using the Connectivity Map
approach (see Kohonen et al., 2014 as an example). In this example (on
Doxorubicin) they have used the top 100 up- and down-regulated genes,
allowing for connectivity mapping to genomic profiles of other agents
with similar modes-of-action.
c. Build work flow (procedure) for such analysis, which could be run for
other compounds starting from the transcriptomics profile.
23 September 2016
7
1. IDENTIFICATION OF RELEVANT PATHWAYS
KEGG pathway enrichment analysis
using InCroMAP
Omics data from in vitro studies
• HepaRG (human hepatoma-derived cells),
• HepG2 (hepatocellular carcinoma-derived
cell line), and
• hES-DE-Hep (hepatocyte-like cells derived
from embryonic stem cells)
Relevant pathways were identified
using KEGG pathway enrichment
approach
23 September 2016
8
1. IDENTIFICATION OF RELEVANT PATHWAYS
Most relevant pathways
identified in HepaRG
cells (A), HepG2 (B) and
hESC DE-Hep cells (C)
23 September 2016
9
1. IDENTIFICATION OF RELEVANT PATHWAYS
• The analysis of HepaRG cells confirmed the interaction of PBO with cytochrome
P450 but this was not observed in HepG2 and hES-DE-Hep cells
• The pathways identified in HepaRG cells showed concentration- and timedependent enrichment, as the treatment with the higher concentration showed
most significant effects at 72h comparative with 24h
• A similar effect was observed in HepG2 cells but the pathways enriched were
different. In the case of of hES-DE-Hep a cell batch-dependent response was seen
• Top pathways enriched in at least in two cell lines were selected and analyzed in CTD
Enriched pathway (Q<0.05) KEGG ID
ECM-receptor interaction
04512
Focal adhesion
04510
PI3K-Akt signaling pathway
04151
Adherens junction
04520
Cell cycle
04110
DNA replication
03030
Proteoglycans in cancer
05205
HepaRG
*
*
*
HepG2
*
*
*
*
23 September 2016
hESC DE-Hep
*
*
*
*
*
*
*
10
2. CORRELATION BETWEEN PATHWAYS AND DISEASES
Analysis of top enriched
pathways in all cell lines
• The AO Fibrosis, appears on
rank 32 of the list of diseases
associated with the identified
pathways
Number of
Rank Disease Name
Disease ID
associated
genes
1
Dermatitis,
Allergic MESH:D017449
100
Contact
• Genes of three pathways are
associated with this disease these genes are mainly
related to pathways identified
with the HepaRG cells
2
Prostatic Neoplasms
MESH:D011471
49
3
Breast Neoplasms
MESH:D001943
48
4
Stomach Neoplasms
MESH:D013274
44
5
Lung Neoplasms
MESH:D008175
40
Analysis of top enriched
pathways on separate cell
models
32
Fibrosis
MESH:D005355
15
• 21 genes associated with
fibrosis in the HepaRG cells
(table), comparative with 4
genes associated with fibrosis
in HepG2 cells
Number of
Rank
Disease Name
Disease ID
associated
genes
4
Drug-Induced
Liver MESH:D056486
50
MESH:D005355
21
Injury
43
Fibrosis
23 September 2016
11
5. VERIFYING ADVERSE EFFECTS BY TESTING FOR
SPECIFIC KEY EVENTS
KE stellate cell activation
for which data does exist
(unpublished data)
AOPKB describes three
different methods for testing
KE collagen accumulation
AO Fibrosis is a potential
adverse outcome for
exposure to PBO
Source: http://aopkb.org/
KE TGF-β1 expression is measured by ELISA but no data for PBO was available
At the transcription level, no effect was observed (however, the AOP does not
list gene expression as a valid test for this KE)
23 September 2016
12