BioNetVisA 2018 ECCB workshop
From biological network reconstruction to data visualization and analysis in molecular biology and medicine.
http://eccb18.org/workshop-2/
https://bionetvisa.github.io/
2. Why disease maps?
• Conventional single-gene biomarkers have a demonstrated clinical utility.
However, their success is purely probabilistic, often modest and frequently
lack any mechanistic anchoring to the fundamental cellular processes
responsible for the disease or therapeutic response.
• Despite most biomarkers used are single gene variants, most human genetic
diseases (and almost all traits) have a modular nature: Causative genes for
the same or phenotypically similar diseases may generally reside in the same
biological module, either a protein complex (Lage et al, 2007), a sub-
network of protein interactions (Lim et al, 2006) , or a pathway (Wood et al,
2007)
Goh 2007 PNAS
Same disease
in different
populations is
caused by
different
genes
affecting the
same
functionsFernandez, 2013, Orphanet J Rare Dis.
Disease
genes are
close in the
interactome
3. Pathways: maps of cell activity
(in sickness and in health)
disease-maps.org/
www.genome.jp/kegg
reactome.org www.wikipathways.org/
Oncogénesis
Alzheimer
More disease maps…
Parkinson
https://www.pathwaycommons.org/
5. Defining pathway activity
We first need a map: pathways are defined in different repositories (KEGG,
Reactome, Biocarta, disease maps, etc.)
What pathway level makes a real biological meaning?
Gene sub-pathway pathway
Enrichment methods
(pathway-level): Different
and often opposite cell
behaviors are triggered by
the same pathway.
E.g.: death and survival
Gene level: The same gene can trigger different (and
often opposite) responses, depending on the stimulus
Survival
Death
Sub-pathway
(elementary circuit)
connects stimulus to
response
7. How realistic are models of
pathway activity?
Beyond static biomarkers—The activity
of signalling networks as an alternate
biomarker?
Fey et al., Sci. Signal. 8, ra130 (2015).
Inability of JNK activation (that mediates
apoptosis) is associated to bad prognostic,
irrespective of MYCN amplification status
Problem:
ODE can
efficiently
solve only
small
systems Construct, activity inferred
The system is more than the sum of its parts!
8. 𝑆 𝑛 = 𝜐 𝑛 ∙ 1 − 1 − 𝑠 𝑎
𝑠 𝑎 ∈𝐴
⋅ 1 − 𝑠𝑖
𝑠𝑖∈𝐼
From individual gene
expression profiles…
…to profiles of circuit
activity (and
functional activity)
Two types of activities
Signal propagation models of
signaling pathways
Are scalable
10. Signaling activity trigger cell functions
directly related to cancer progression
DNA replication function is a construct: the activity is inferred not measured
DNA replication= f(gene1, gene2, … genen) Hidalgo et al., 2017 Oncotarget
11. Actually, signal activity triggers
cancer hallmarks
Hanahan, Weinberg, 2011
Hallmarks of cancer: the next
generation. Cell 144, 646
Negative regulation of release of cytochrome c
from mitochondria (inhibition of apoptosis)
12. The inferred function activity (mechanistic
biomarker) is more correlated to survival
than the activity of any gene (conventional
biomarker) in the circuit
p-val=5.9x10-8
The system is more than the sum of its parts!
13. Actionable models
The real advantage of models is that, the same way they can be used to
transforming omics data into measurements of cell functionalities that
help in the interpretation of disease mechanisms and drug MoA, they can
be used to test mechanistic hypothesis such as “what if I suppress (or
over-express) this (these) gen(es)?” This lead to the concept of
actionable models.
By simulating changes of gene expression/activity in a given system it is
easy to:
• Directly study of the consequences of induced gene over-expressions
or KOs, inhibitions, drugs, etc., alone or in combinations.
• Carry out reverse engineering studies about what genes need to be
perturbed to change specific cell functionalities, such as:
• Reverting the “normal” functional status of a cell
• Selectively kill diseased cells without affecting normal cells
• Enhancing or reducing cell functionalities (e.g., apoptosis or
proliferation, respectively, to fight cancer)
• Etc.
14. Model validation (1)
The activity of some signaling circuits is correlated with cell survival
Increasedsurvival
Increased circuit activity
Repression of apoptosis
Activation of proliferation
Functions:
Activation of apoptosis
Repression of proliferation
Survival data from Achilles cell line KOs (Broad Institute) can be
compared to the change in circuit activities predicted by the model
Essential circuits: once found, other ways of deactivating these circuits
can be find, opening the door to knowledge-based target discovery
Onco-circuit Tumor suppressor circuit
15. Model validation (2)
Prediction of other gene targets, whose inhibition
(modeled KO) deactivate these circuits and
consequently decrease cell viability
1 04 9
8
7
5 2 6
3 1 2
3Achilles I
Circuit activity
with gene KO
Circuit activity
Cellsurvival
4
5
9
0
6
7
8
Potential target (inhibition of circuit activity)
16. Model validation (3)
Deactivation of tumor
suppressor modules
Deactivation in onco-modules
Prostate cell line Skin cell line
Validation of the real KO effects with Achilles II (Tsherniak A, et al. 2017, Cell)
We obtained 65% correct predictions (Çubuk et al., 2018, Cancer Res)
18. Actionable pathway models
Transcription
We can inhibit EGFR (target of Afanatib) by
reducing its activity value (0.56 in cancer).
Absolute KO value = 0
Estrogen signaling pathway http://pathact.babelomics.org/
19. Actionable pathway models
The inhibition of the transcription sought has been attained, but six more pathways
have been affected in different ways
http://pathact.babelomics.org/
20. Actionable pathway models
Cell cycle inhibited in
Proteoglycans in cancer pathway
Transcription, angiogenesis and other
are inhibited in ErbB signaling pathway
Cell cycle is inhibited in Oxytocin
signaling pathway
http://pathact.babelomics.org/
21. Simulating drug inhibition
“Ideal” KO of
EGFR affects 7
pathways
Real inhibition with
Afanatib affects 11
pathways
Inhibition with
broader spectrum
Trastuzumab
affects 13 pathways
22. There are 94 modules that recapitulate the
main aspects of metabolism of lipids,
carbohydrates, amino acids and nucleotides
Metabolic
activity:
Differential metabolic activity: two conditions are compared by means of a
Wilcoxon test (FDR adjusted across modules)
Simple propagation models can also be
extended to model metabolic pathways
24. Metabolic modules also capture cell
functionality associated to cancer prognostic
High activity of Guanine ribonucleotide biosynthesis and Pyrimidine
ribonucleotide biosynthesis modules is associated to low survival.
These modules are target of Mercaptopurine and Gemcitabine.
The mechanism of action of these drugs involves inhibition of DNA synthesis
and that leads to cell death
25. Correspondence between
module activity and change in
metabolite abundance
Module Cancer type First metabolite
Intermediate
metabolite
End
metabolite
Prediction
M00035_1
C00073
C00019 /
C00021 /
C00155
C02291
BRCA 1,9 NA / 2.7 / NA 7,8 Up Correct
KIRC 0,5 0.9 / 1.7 / 0.6 NA Down NA
M00100
C00319 C06124 C00346
BRCA 11,1 NA 17,3 Up Correct
KIRC 2,2 NA 0,8 Down Correct
M00135
C00134
C02714 /
C05936 /
C02946
C00334
BRCA 12,9 NA / NA / NA 1,2 Down Correct
KIRC 3,6 NA / NA / NA 0,3 Down Correct
26. The metabolizer web server
http://metabolizer.babelomics.org/
Input: gene expression matrix
Analyses:
• Differential metabolic activity
• Case control
• Correlation
• KnockOut.
• Tests the effect of a KO in
the metabolic profile
• Auto KO: in a case/control
search for the KO that
makes the metabolic activity
profile of cases as close as
possible to thoso of controls
(phenotype reversion)
• Prediction: Builds a class
predictor (RF or SVM) based on
metabolic profiles
27. An example of KO prediction
Pyrimidine degradation pathway was predicted to be an onco-module in gastric
cancer cell lines. Predicted genes that switch the pathway off are DPYD, DPYS
(confirmed in Achilles) and UPB1 (not present in Achilles)
28. Prediction of gene essentiality from
metabolic pathway essentiality
UPB1 encodes an enzyme (β-ureidopropionase) that catalyzes the last step in the
pyrimidine degradation pathway, required for epithelial-mesenchymal transition
Pyrimidine degradation pathway was predicted to be an onco-module in gastric
cancer cell lines. Predicted genes that switch the pathway off are DPYD, DPYS
(confirmed in Achilles) and UPB1
29. Models of cell functional activity bring
the dream of precision personalized
(even individualized) treatments closer
From: Dopazo, 2014, Genomics and transcriptomics in drug discovery. Drug Discovery Today
30. Clinical Bioinformatics Area
Fundación Progreso y Salud, Sevilla, Spain, and…
...the INB-ELIXIR-ES, National Institute of Bioinformatics
and the BiER (CIBERER Network of Centers for Research in Rare Diseases)
@xdopazo
@ClinicalBioinfo
Follow us on
twitter
https://www.slideshare.net/xdopazo/