DECIPHERING CANCER MECHANISMS BY INTEGRATIVE NETWORK ANALYSIS - - PowerPoint PPT Presentation

▶
deciphering cancer mechanisms by integrative network
SMART_READER_LITE
LIVE PREVIEW

DECIPHERING CANCER MECHANISMS BY INTEGRATIVE NETWORK ANALYSIS - - PowerPoint PPT Presentation

Feb 18, 2024 290 likes •875 views

DECIPHERING CANCER MECHANISMS BY INTEGRATIVE NETWORK ANALYSIS Research Seminar Duke-NUS Medical School June 2014 Sriganesh Srihari Institute for Molecular Bioscience, The University of Queensland, QLD, Australia Cancer: A large class of

slide-1
SLIDE 1

DECIPHERING CANCER MECHANISMS BY INTEGRATIVE NETWORK ANALYSIS

Research Seminar Duke-NUS Medical School June 2014

Sriganesh Srihari

Institute for Molecular Bioscience, The University of Queensland, QLD, Australia

slide-2
SLIDE 2

Cancer: A large class of diseases affecting different organs of the body

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

2

Sources: National Cancer Institute USA; Science

Even if it affects the same

  • rgan site
  • E.g. Breast cancer – five

“intrinsic” subtypes (identified from gene-expression; Perou et al. 2000)

  • At least ten subtypes from genomic

and expression data (Curtis et al., 2012)

slide-3
SLIDE 3

Cancer: The origin is in the genome…

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

5

Hampton et al., Genome Research 2009

slide-4
SLIDE 4

Leading up to pathways and processes

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

6

Sources: Nature Reviews, Science

slide-5
SLIDE 5

Understanding dysregulation in pathways:

Usually studied as individual genes

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland 7

#Up-regulated = 8 #Down-regulated = 8 So, is this pathway up- or down-regulated?

  • Mean / maximum / voting of genes?

“Top” portion is down, but “bottom” portion is up-regulated.

  • Where do you draw the boundaries?

By studying genes individually, we are missing their aggregate effect.

  • Not taking into account the structure or topology of the pathway.
slide-6
SLIDE 6

We obtain an aggregate or “systems biology” view of underlying mechanisms.

Understanding dysregulation in pathways:

Studying complexes

8

Proteins seldom perform their functions in isolation, but instead form stable functional complexes. By looking at complexes,

  • Aggregate the effect of individual genes;
  • Factor in topological structure.

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

slide-7
SLIDE 7

Complexes in Pathways and Processes Affected in Cancer

9

Sources: Nature Reviews, Science

slide-8
SLIDE 8

Complexes in Diseases

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

10

slide-9
SLIDE 9

ANALYSING COMPLEXES IN CANCER

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

11

slide-10
SLIDE 10

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

12

Human complex databases: CORUM (Mewes et al., NAR 2004) Havugimana et al. (Cell, 2012) Coverage ~30 - 40%

slide-11
SLIDE 11

Complex Identification from Protein Interactions

A typical pipeline (Spirin & Merny, PNAS 2003)

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

13

  • 1. PPI network
  • 2. Complexes
  • 3. Validation

Low- and high- throughput experiments. Assembled as a network after filtering noise. Clustering the PPI network to predict complexes.

Validate against bona fide complexes. Study roles of novel complexes.

Human complex databases: CORUM (Mewes et al., NAR 2004) Havugimana et al. (Cell, 2012) Coverage ~30 - 40%

slide-12
SLIDE 12

Complex Prediction from Protein Interaction Networks

Computational methods in the literature (Srihari & Leong, 2013)

14 There are several methods for identifying complexes from PPI networks. Experiments on yeast suggest ~75% coverage. Srihari S and Leong HW, J Bioinf Comp Biol 2013.

slide-13
SLIDE 13

Can we use this pipeline ?

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

15 “Cancer PPI network” Complexes in cancer Validate against known, study roles of novel complexes in cancer

slide-14
SLIDE 14

Can we use this pipeline ?

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

16

  • But,
  • how do we gather such a “cancer PPI network” ?
  • what constitutes the network?
  • which complexes are affected in cancer?
  • not all complexes are involved in cancer.

‘Cancer PPI network’ Complexes in cancer

PPI networks do not have contextual information Can we predict ‘cancer PPI network’?

Validate against known, study roles of novel complexes in cancer

Sure!

slide-15
SLIDE 15

Identify complexes that are “dysfunctional” in cancer

  • Track complexes for behavioural differences across conditions

by integrating diverse information,

  • Mutated genes coding for dysfunctional proteins within complexes
  • Changes in expression of coding genes
  • Changes in protein composition or abundance

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

17 The key! mRNA expression Mutations in genes or chromosomal locii

Normal tumour

Possible to predict

slide-16
SLIDE 16
  • Integrate the following two kinds of data:
  • PPI network (BioGrid v3.1.93; Stark et al. 2011)

5824 proteins 29600 interactions [high-quality post filtering]

  • Gene expression

39 matched normal-tumour samples from pancreatic adenocarcinoma patients (Badea et al. 2008)

  • Extract complexes from the PPI network using clique-merging (Srihari et al.,

2013) Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

18

(Why) Does this help?

A case study mapping gene expression from Normal and Tumour pancreatic conditions onto PPIs and complexes

Mapping gene co-expression [-1,+1]

slide-17
SLIDE 17

(Why) Does this help?

A case study mapping gene expression from Normal and tumour pancreatic conditions onto PPIs and complexes

19 Significant loss in correlations of PPIs – “accelerators” as well as “brakes”. Significant loss in correlations for complexes.

(KS test: 23.11 > K0.05 = 1.36) (KS test: 1.69 > K0.05 = 1.36)

The aggregate effect of using protein pairs or complexes!

slide-18
SLIDE 18

Affected Protein Interactions

A case study “mapping” gene expression from Normal and tumour pancreatic conditions onto PPIs and complexes

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

20

RBPMS--RHOXF2 SMN1--TMSB4X TGFB1 NFKB

“Jumps” in correlation of PPIs from normal to tumour

KRAS

RHOXF2(PEPP2) Involved in carcinogenesis in gastric and pancreatic cell lines.

(Shibata-Minoshima et al., 2012)

(Exp mean 4.57, 4.63)

slide-19
SLIDE 19

CONTOUR: An Enhanced Pipeline to Detect Dysfunctional Complexes in Cancer

Integrating PPI, Gene Expression and Mutation datasets

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

21

+

1 2 s 1 2 t tumour Normal

Conditional PPI networks

Identify and match complexes between conditions Normal tumour Gene expression, mutation profiles

Generic PPI network

Dysfunctional complexes

Srihari S & Ragan MA, Bioinformatics 2013

slide-20
SLIDE 20

APPLICATION TO CANCERS

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

22

slide-21
SLIDE 21

Cancer Conditions Studied

  • Normal vs Pancreatic adenocarcinoma (PDAC)
  • PDAC – constitutes 95% of pancreatic tumours
  • 39 pairs of matched gene expression samples from Badea et al.

(2008) from normal and PDAC tissues

  • Mutation profiles of 1169 genes from Jones et al. (Science, 2008)
  • BRCA1 tumours vs BRCA2 tumours
  • Germline defects in BRCA1 and BRCA2 genes
  • Deficient in homologous recombination-based DSB repair
  • Profiling of familial breast tumours (kConfab consortium)
  • Expression data from Waddell et al. (2010)

23

BRCA1 tumours ~ basal-like

  • Aggressive
  • Mostly triple-negative (ER/PR/HER2-)

BRCA2 tumours ~ luminal or more heterogeneous (luminal and HER2+)

  • Less aggressive (at least luminal-A)

Badea et al. Hepatogastroenterology 2008 Jones et al. Science 2008 Waddell et al. Breast Cancer Treat Res 2010

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

slide-22
SLIDE 22

Differences between conditional PPI networks

24

(KS test: 23.11 > K0.05 = 1.36) (KS test: 22.85 > K0.05 = 1.36)

Top enriched GO terms for interactions showing ≥ |1.0| change: Cell cycle, chromatin organization, DNA repair and RNA splicing. Pancreatic: KRAS, TGF, RAD21, STAT1, STAT3, P53, SMAD4. Breast: BRCA1, BRCA2, TP53, BRE, BRCC3.

slide-23
SLIDE 23

Differential PPI Network: Normal vs PDAC

25

High cut-off: |1.0| Proteins: 558 Interactions: 519 (very sparse) Interactions PDAC vis-à-vis Normal Red: Weakened Green: Strengthened

Top enriched GO terms for interactions showing ≥ |1.0| change: Cell cycle, chromatin organization, DNA repair and RNA splicing.

EP300 PLK1 ANXA2 PELP1

slide-24
SLIDE 24

Differential PPI Network: Normal vs PDAC

26

slide-25
SLIDE 25

Changes in correlation of complexes

Normal vs PDAC and BRCA1 vs BRCA2 tumours

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

27

(KS test: 1.69 > K0.05 = 1.36) (KS test: 5.48 > K0.05 = 1.36)

CORUM

slide-26
SLIDE 26

Changes in correlation of complexes:

Normal vs PDAC and BRCA1 vs BRCA2

28

Correlation of complexes Correlation of complexes Category Condition #Complexes Max Avg Condition #Complexes Max Avg Our Normal 1.206 0.292 BRCA1 0.863 0.218 256 277 PDAC 0.757 0.154 BRCA2 0.479 0.027 Correlation of complexes Correlation of complexes Category Condition #Complexes Max Avg Condition #Complexes Max Avg CORUM Normal 1.037 0.216 BRCA1 0.702 0.188 189 441 PDAC 0.448 0.113 BRCA2 0.512 0.059

Overall loss in correlation of complexes in PDAC vis-à-vis Normal and BRCA2 vis-à-vis BRCA1.

slide-27
SLIDE 27

Functional enrichment analysis of dysfunctional complexes (Gene Ontology)

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

29

Go category GO term % genes p-value Cell cycle 4.60 3.50E-13 Pathways in cancer 6.20 5.80E-07 RIG-I-like signalling 2.20 1.10E-05 Neurotrophin signalling 3.00 1.70E-05 KEGG pathways Nucleotide exicision repair 1.70 1.90E-05 Pancreatic cancer 2.10 5.70E-05 Adipocytokine signalling 2.00 9.70E-05 Regulation of autophagy 1.30 3.40E-04 Mismatch repair 1.00 5.20E-04 Wnt signalling 2.80 2.20E-03 Cell cycle 17.30 1.60E-35 Biological Chromosome organization 13.00 6.20E-33 Process RNA splicing 9.20 2.50E-28 Chromatin modification 8.90 1.00E-27

Normal vs PDAC

Go category GO term % genes p-value Cell cycle 3.20 2.70E-07 Pathways in cancer 5.80 2.90E-07 Nucleotide exicision repair 1.60 1.50E-05 DNA replication 1.40 6.40E-05 KEGG pathways Adipocytokine signalling 1.80 7.50E-07 Apoptosis 2.10 1.20E-04 Neurotrophin signalling 2.30 9.50E-04 Homologous recombination 1.00 1.60E-03 Insulin signalling 2.20 6.00E-03 Mismatch repair 0.90 2.80E-03 Chromosome organization 14.30 1.50E-43 Biological Chromatin organization 12.20 1.30E-40 Process Cell cycle 14.50 7.00E-25 Regulation of transcription 31.60 1.10E-24

BRCA1 tumour vs BRCA2 tumour

~ 500 genes in each case

slide-28
SLIDE 28

Functional enrichment analysis of dysfunctional complexes (Gene Ontology)

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

30

Go category GO term % genes p-value Cell cycle 4.60 3.50E-13 Pathways in cancer 6.20 5.80E-07 RIG-I-like signalling 2.20 1.10E-05 Neurotrophin signalling 3.00 1.70E-05 KEGG pathways Nucleotide exicision repair 1.70 1.90E-05 Pancreatic cancer 2.10 5.70E-05 Adipocytokine signalling 2.00 9.70E-05 Regulation of autophagy 1.30 3.40E-04 Mismatch repair 1.00 5.20E-04 Wnt signalling 2.80 2.20E-03 Cell cycle 17.30 1.60E-35 Biological Chromosome organization 13.00 6.20E-33 Process RNA splicing 9.20 2.50E-28 Chromatin modification 8.90 1.00E-27

Normal vs PDAC

Go category GO term % genes p-value Cell cycle 3.20 2.70E-07 Pathways in cancer 5.80 2.90E-07 Nucleotide exicision repair 1.60 1.50E-05 DNA replication 1.40 6.40E-05 KEGG pathways Adipocytokine signalling 1.80 7.50E-07 Apoptosis 2.10 1.20E-04 Neurotrophin signalling 2.30 9.50E-04 Homologous recombination 1.00 1.60E-03 Insulin signalling 2.20 6.00E-03 Mismatch repair 0.90 2.80E-03 Chromosome organization 14.30 1.50E-43 Biological Chromatin organization 12.20 1.30E-40 Process Cell cycle 14.50 7.00E-25 Regulation of transcription 31.60 1.10E-24

BRCA1 tumour vs BRCA2 tumour

Genome-stability mechanisms – DNA damage response and cell cycle severely affected in tumours. Brought about by significant rewiring of physical interactions leading up to dysfunctioning of complexes.

~ 500 genes in each case

slide-29
SLIDE 29

Break-up of complexes based on correlation in the two tumours

31

Change (increase and decrease) in correlation of complexes Normal vs PDAC BRCA1 vs BRCA2 #Complexes Max Avg #Complexes Max Avg M' 159 0.969 0.336 225 0.761 0.281 M" 96 0.421 0.192 52 0.543 0.197 Correlation of complexes Correlation of complexes Condition #Complexes Max Avg Condition #Complexes Max Avg Normal 1.206 0.292 BRCA1 0.863 0.218 256 277 PDAC 0.757 0.154 BRCA2 0.479 0.027

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

slide-30
SLIDE 30

Break-up of complexes based on correlation in the two tumours

32

Change (increase and decrease) in correlation of complexes Normal vs PDAC BRCA1 vs BRCA2 #Complexes Max Avg #Complexes Max Avg M' 159 0.969 0.336 225 0.761 0.281 M" 96 0.421 0.192 52 0.543 0.197 Correlation of complexes Correlation of complexes Condition #Complexes Max Avg Condition #Complexes Max Avg Normal 1.206 0.292 BRCA1 0.863 0.218 256 277 PDAC 0.757 0.154 BRCA2 0.479 0.027

Compensatory mechanisms coming into play in tumours

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

slide-31
SLIDE 31

Hallmarks of Cancer

Hanahan D, Weinberg RA (Cell 2000)

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

33

Sources: Hanahan and Weinberg, Cell 2000

Increase in correlation of complexes a sign of these?

slide-32
SLIDE 32

Examples of dysfunctional complexes in PDAC

34

NFkB regulates several anti-apoptotic genes (TRAF1 and TRAF2) and checks caspase family of apoptotic genes. Commonly blocked to stop proliferating cells. But recent studies have also shown that NFkB activity sensitises cells to apoptosis and senescence through Fas (Liu et al, J Biol Chem, 2012).

slide-33
SLIDE 33

Examples of dysfunctional complexes in BRCA1 and BRCA2 tumours

35

Higher cell-division rates in BRCA1 tumours? Due to the parallel roles of BRCA1 in mismatch repair?

slide-34
SLIDE 34

Some synergy observed upon siRNA-mediated depletion

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland 36

Preliminary results (more experiments underway):

Combined depletion of RAD51/SMC3 or RAD51/RFC3 shows synergistic killing of up to 70% cancer cells

Cohesin and replication- factor complexes

SUM 159PT ER-/PR- ERRB2+

  • Aggressive cell

lines with RAD51

  • verexpressed
  • HRAS, PIK3CA

mutations

slide-35
SLIDE 35

WHAT IS THE CONTEXT?

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

37

Why are we studying this?

slide-36
SLIDE 36

Synthetic lethality: a new promise for selective killing of cancer cells

38

Targeting the synthetic lethal partners of lost genes (tumour suppressors) can selectively kill cancer cells.

A buffering mechanism in cells

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

slide-37
SLIDE 37

SL-based therapy: how does it work?

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland 39

Targeting compensatory pathways

Liu & Srihari et al., Nucleic Acids Research 2014 (Review)

slide-38
SLIDE 38

Goal: Identifying novel SL relationships

  • Targeting frequently altered genes (drivers)
  • “Oncogene-addiction” (Weinstein & Joe 2002, 2008)
  • Many of the drivers are in fact essential genes,

difficult to target without killing normal cells

  • E.g. KRAS – highly prevalent oncogene (pancreas, lung, colon)
  • Further cannot target tumour suppressors (except during

relapse/resistance)

  • “Genetic dependency” – e.g. PTEN loss -- dependency
  • n dysregulated PI3K kinase activation
  • So target their SL partners

40

Oncogene

(KRAS – TBK1)

Tumour suppressor

(BRCA1 – PARP1)

SL partner

slide-39
SLIDE 39

BRCA1-PARP1: a clinically relevant relationship

  • BRCA1/BRCA2-deficient tumours are sensitive to

PARP inhibition

  • Such cancer cells die upon PARP inhibition
  • In general, PARP is SL with several other HR genes
  • Explored as a therapeutic target in “BRCAness” tumours

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland 41

BRCA1/2

  • HR-mediated DSB repair
  • Replication-fork restart

PARP1

  • A-NHEJ DSB repair
  • Replication-fork

restart

  • Coop with Ku70/80

Compensatory relationship

Liu & Srihari et al., Nucleic Acids Research 2014 (Review)

slide-40
SLIDE 40

BRCA1-PARP1: a clinically relevant relationship

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland 42

  • PARP1-/- mice are viable and fertile; PARP1-ko is not lethal.
  • Cancer cells undergo significant replicative stress
  • Generates considerable lesions through replication fork stalling
  • PARP is required to facilitate fork restart and enable HR-dependent

repair of lesions

  • In the event of deficient HR, further inhibition of PARP is lethal to cancer cells

BRCA1/2

  • HR-mediated DSB repair
  • Replication-fork restart

PARP1

  • A-NHEJ DSB repair
  • Replication-fork

restart

  • Coop with Ku70/80

Compensatory relationship

slide-41
SLIDE 41

Try systems-biology modelling (computational)

  • Look at networks – understand the wiring of functional

dependencies

  • Modelling key signalling pathways (oncogenic, DNA-

damage repair, apoptotic, cell-cycle control) as logic circuits

  • Requires a comprehensive

Knowledge map of pathways

  • Current knowledge inadequate
  • Nevertheless, try this
  • Analysis of pathway networks

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland 43

slide-42
SLIDE 42

We obtain a better aggregate or “systems biology” view of underlying mechanisms.

Understanding dysregulation in pathways: Studying complexes

44

Proteins seldom perform their functions in isolation, but instead form stable functional complexes. By looking at complexes,

  • Aggregate the effect of individual genes;
  • Factor in topological structure.
slide-43
SLIDE 43

Some synergy observed upon siRNA-mediated depletion

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland 45

Preliminary results (more experiments underway):

Combined depletion of RAD51/SMC3 or RAD51/RFC3 shows synergistic killing of up to 70% cancer cells

Cohesin and replication- factor complexes

SUM 159PT ER-/PR- ERRB2+

  • Aggressive cell

lines with RAD51

  • verexpressed
  • HRAS, PIK3CA

mutations

slide-44
SLIDE 44

PARALLEL / RELATED PROJECTS

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland 46

slide-45
SLIDE 45

Other projects on breast cancer

  • Curation of DDR pathways from literature
  • Mechanistic knowledge on DDR pathways (NAR 2014)
  • A database of these pathways to be released soon
  • Evaluation of oncogenic and tumour suppressor/pro-

apoptotic pathways in breast cancer classification

  • Molecular classification & prognostic ability (Briefings

in Bioinf 2014)

  • Developing Boolean logic models for time-series

evaluation

  • E.g. Drug response tracked over time (IEEE/ACM Trans 2013)

47 Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

slide-46
SLIDE 46

CONTOUR

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

48

Srihari and Ragan (2013) Systematic tracking of dysregulated modules identifies novel genes in cancer, Bioinformatics.

slide-47
SLIDE 47

The CONTOUR Workflow

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

49

+

1 2 s 1 2 t tumour Normal Conditional PPI networks Identify and match complexes between conditions Normal tumour Gene expression, mutation profiles Generic PPI network Dysfunctional complexes

Step 1 Step 2 + 3

slide-48
SLIDE 48

The CONTOUR Workflow

Step 1: Integrating PPI, gene expression and mutation data

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

50

+

1 2 s 1 2 t tumour Normal Conditional PPI networks Identify and match complexes between conditions Normal tumour Gene expression, mutation profiles Generic PPI network Dysfunctional complexes

Step 1 Step 2 + 3

slide-49
SLIDE 49

The CONTOUR Workflow

Step 1: Integrating PPI, gene expression and mutation data

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

51

+

tumour Normal Gene expression, mutation profiles Generic PPI network

Step 1

Backbone: PPI network

  • H = (VH,EH): Human generic PPI network
  • w(p,q): weight on each edge e =(p,q)  EH

Evidence: gene expression

  • E = set of all protein (gene) pairs
  • Co-expression for all pairs in EH
  • Co-expression for all pairs in E \ EH
  • Frequency distribution for EH
  • Frequency distribution for E \ EH

Normal condition

X

X’

Positive interaction set Negative interaction set

(p,q)  [ x, x + 0.10)

|X| |X| + |X ’|

Pg [ (p,q)  EH ] =

slide-50
SLIDE 50

The CONTOUR Workflow

Step 1: Integrating PPI, gene expression and mutation data

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

52

+

tumour Normal Gene expression, mutation profiles Generic PPI network

Step 1

Backbone: PPI network

  • H = (VH,EH): Human generic PPI network
  • w(p,q): weight on each edge e =(p,q)  EH

Evidence: mutation profiles

  • E = set of all protein (gene) pairs
  • tumourigenity likelihood for pairs in EH
  • tumourigenity likelihood for pairs in E \ EH
  • Frequency distribution for EH
  • Frequency distribution for E \ EH

Normal condition

Y

Y’

Positive interaction set Negative interaction set

(p,q)  [ y, y + 0.10)

|Y| |Y| + |Y ’|

Pm [ (p,q)  EH ] =

slide-51
SLIDE 51

The CONTOUR Workflow

Step 1: Integrating PPI, gene expression and mutation data

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

53

Backbone: PPI network

  • H = (VH,EH): Human generic PPI network
  • w(p,q): weight on each edge e =(p,q)  EH

Combining the two evidence Gene expression: Pg Mutation profiles: Pm wN(p,q) = w(p,q) .

  • Pg. Pm

Normal condition

  • Pg. Pm + (1 – Pg ) * (1 – Pm )

+

Normal Normal Generic PPI network Gene expression, mutation profiles

Normal PPI network Re-weight the interactions in the PPI network using gene expression and mutation evidence. We do not change the topology of the network.

slide-52
SLIDE 52

The CONTOUR Workflow

Step 1: Integrating PPI, gene expression and mutation data

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

54

Backbone: PPI network

  • H = (VH,EH): Human generic PPI network
  • w(p,q): weight on each edge e =(p,q)  EH

Combining the two evidence Gene expression: Pg Mutation profiles: Pm wT(p,q) = w(p,q) .

  • Pg. Pm

tumour condition

  • Pg. Pm + (1 – Pg ) * (1 – Pm )

+

tumour Conditional PPI networks tumour Gene expression, mutation profiles

tumour PPI network Re-weight the interactions in the PPI network using gene expression and mutation evidence. We do not change the topology of the network.

slide-53
SLIDE 53

The CONTOUR Workflow

Step 2: Mining complexes from the conditional PPI networks

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

55

+

1 2 s 1 2 t tumour Normal Conditional PPI networks Identify and match complexes between conditions Normal tumour Gene expression, mutation profiles Generic PPI network Dysfunctional complexes

Step 1 Step 2 + 3

slide-54
SLIDE 54

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

56

The CONTOUR Workflow

Step 2: Mining complexes from the conditional PPI networks

tumour Normal Similar to CMC (Liu et al., Bioinformatics 2009)

slide-55
SLIDE 55

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

57

The CONTOUR Workflow

Step 3: Matching complexes between conditions

1 2 n 1 2 m J (N1,T1) N T J (N1,T2) J (N2,T2) J (Nn, Tm) ( Take only the best matching Tj for a Ni ) J (N1,T1) > J (N1,T2)

Matching

  • Satisfy two requirements,
  • similarity in protein composition
  • a non-zero change in co-expression density

between the conditions Co-expression density

  • Co-expression density of a complex C

in a condition

CoExp(p,q) values are Fisher-transformed

Density represents the ‘activity’ of the complex.

  • High co-expression means the

complex is functional in that condition Maximal weighted matching problem

slide-56
SLIDE 56

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

58

The CONTOUR Workflow

Step 3: Matching complexes between conditions

Merge Discard Merge Discard Match

1 2 3 Normal tumour Ranking

  • f cliques

Swapped genes

Complex Complex

1 2 n 1 2 m J (N1,T1) N T J (N1,T2) J (N2,T2) J (Nn, Tm) ( Take only the best matching Tj for a Ni ) Highly indicative of involvement in cancer.

slide-57
SLIDE 57

Summary

  • CONTOUR (v1) – Tracking of dysregulated

complexes

  • Integrates PPI, gene expression and mutation data
  • Dysregulated complexes harbor cancer genes and tumour

suppressors

  • Indicative of compensatory mechanisms coming into play
  • CONTOUR (v2) – Identifying influential TFs

controlling dysregulated complexes

  • Influential TFs work in cooperative and counteractive ways
  • Tightly associated co-regulation

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

59

slide-58
SLIDE 58

Thank You…

Sriganesh Srihari, Institute for Molecular Bioscience, The University of Queensland

60

Acknowledgements Institute for Molecular Bioscience, UQ

Prof Mark Ragan and his group

UQ Centre for Clinical Research

Dr Peter Simpson

Queensland Institute of Medical Research

Prof Kum Kum Khanna and her group