Stephen Friend, Nov 1, 2011. Institute for Cancer Research, Oslo, Norway

1. Actionable Cancer Network Models
And Open Medical Information Systems
Integrating layers of omics data models and compute spaces
Stephen Friend MD PhD
Sage Bionetworks (Non-Profit Organization)
Seattle/ Beijing/ Amsterdam
ICR Oslo
November 1, 2011

5. Why not use data intensive science
to build models of disease
Current Reward Structures
Organizational Structures and Tools
Pilots
Opportunities

6. What is the problem?
• Regulatory hurdles too high?
• Low hanging fruit picked?
• Payers unwilling to pay?
• Genome has not delivered?
• Valley of death?
• Companies not large enough to execute on strategy?
• Internal research costs too high?
• Clinical trials in developed countries too expensive?
In fact, all are true but none is the real problem

7. What
is
the
problem?
We
need
to
rebuild
the
drug
discovery
process
so
that
we
be6er
understand
disease
biology
before
tes8ng
proprietary
compounds
on
sick
pa8ents

8. What
is
the
problem?
Most
approved
cancer
therapies
assumed
tumor
indica8ons
would
represent
homogenous
popula8ons
Most
new
cancer
therapies
are
in
search
of
single
altered
components
Our
exis8ng
tumor
models
o>en
assume
pathway
knowledge
sufficinet
to
infer
correct
therapies

9. Personalized Medicine 101:
Capturing Single bases pair mutations = ID of responders

10. Reality: Overlapping Pathways

11. The value of appropriate representations/ maps

13. “Data Intensive” Science- Fourth Scientific Paradigm
Equipment capable of generating
massive amounts of data
IT Interoperability
Open Information System
Host evolving computational models
in a “Compute Space”

17. WHY
NOT
USE
“DATA
INTENSIVE”
SCIENCE
TO
BUILD
BETTER
DISEASE
MAPS?

18. what will it take to understand disease?
DNA
RNA
PROTEIN
(dark
maGer)
MOVING
BEYOND
ALTERED
COMPONENT
LISTS

19. 2002 Can one build a “causal” model?

20. How is genomic data used to understand biology?
RNA amplification
Tumors
Microarray hybirdization
Tumors
Gene Index
Standard GWAS Approaches Profiling Approaches
Identifies Causative DNA Variation but Genome scale profiling provide correlates of disease
provides NO mechanism  Many examples BUT what is cause and effect?
 Provide unbiased view of
molecular physiology as it
relates to disease phenotypes
trait
 Insights on mechanism
 Provide causal relationships
and allows predictions
20
Integrated Genetics Approaches

21. Integration of Genotypic, Gene Expression & Trait Data
Schadt et al. Nature Genetics 37: 710 (2005)
Millstein et al. BMC Genetics 10: 23 (2009)
Causal Inference
“Global Coherent Datasets”
• population based
• 100s-1000s individuals
Chen et al. Nature 452:429 (2008) Zhu et al. Cytogenet Genome Res. 105:363 (2004)
Zhang & Horvath. Stat.Appl.Genet.Mol.Biol. 4: article 17 (2005) Zhu et al. PLoS Comput. Biol. 3: e69 (2007)

22. Gene Co-Expression Network Analysis
Define a Gene Co-expression Similarity
Define a Family of Adjacency Functions
Determine the AF Parameters
Define a Measure of Node Distance
Identify Network Modules (Clustering)
Relate the Network Concepts to
External Gene or Sample Information 22
Zhang B, Horvath S. Stat Appl Genet Mol Biol 2005

23. Constructing Co-expression Networks
Start with expression measures for genes most variant genes across 100s ++ samples
1 2 3 4 Note: NOT a gene
expression heatmap
1
1 0.8 0.2 -0.8
Establish a 2D correlation matrix 2
for all gene pairs
expression
0.8 1 0.1 -0.6
3
0.2 0.1 1 -0.1
4
-0.8 -0.6 -0.1 1
Brain sample
Correlation Matrix
Define Threshold
eg >0.6 for edge
1 2 4 3 1 2 3 4
1 1
1 4 1 1 1 0 1 1 0 1
2 2
1 1 1 0 1 1 0 1
1 1 1 0 Hierarchically 3
Identify modules 4 0 0 1 0
2 3 cluster
4
3 0 0 0 1 1 1 0 1
Network Module Clustered Connection Matrix Connection Matrix
sets of genes for which many
pairs interact (relative to the
total number of pairs in that
set)

24. Preliminary Probabalistic Models- Rosetta /Schadt
Networks facilitate direct
identification of genes that are
causal for disease
Evolutionarily tolerated weak spots
Gene symbol Gene name Variance of OFPM Mouse Source
explained by gene model
expression*
Zfp90 Zinc finger protein 90 68% tg Constructed using BAC transgenics
Gas7 Growth arrest specific 7 68% tg Constructed using BAC transgenics
Gpx3 Glutathione peroxidase 3 61% tg Provided by Prof. Oleg
Mirochnitchenko (University of
Medicine and Dentistry at New
Jersey, NJ) [12]
Lactb Lactamase beta 52% tg Constructed using BAC transgenics
Me1 Malic enzyme 1 52% ko Naturally occurring KO
Gyk Glycerol kinase 46% ko Provided by Dr. Katrina Dipple
(UCLA) [13]
Lpl Lipoprotein lipase 46% ko Provided by Dr. Ira Goldberg
(Columbia University, NY) [11]
C3ar1 Complement component 46% ko Purchased from Deltagen, CA
3a receptor 1
Tgfbr2 Transforming growth 39% ko Purchased from Deltagen, CA
Nat Genet (2005) 205:370 factor beta receptor 2

25. Our ability to integrate compound data into our network analyses
db/db mouse
(p~10E(-30))
= up regulated
= down regulated
db/db mouse
(p~10E(-20)
p~10E(-100))
AVANDIA in db/db mouse

26. Genomic
Literature
Protein-­‐Protein
Complexes
Transcriptiona
l
Signaling
THE EVOLUTION OF SYSTEMS BIOLOGY
Mol.
Profiles
Structure
Model
Evolution
Disease
Models
Model
Topology
Physiologic
/
Model
Dynamics
Pathologic
Phenotype
Regulation

27. Extensive Publications now Substantiating Scientific Approach
Probabilistic Causal Bionetwork Models
• >80 Publications from Rosetta Genetics
Metabolic "Genetics of gene expression surveyed in maize, mouse and man." Nature. (2003)
Disease "Variations in DNA elucidate molecular networks that cause disease." Nature. (2008)
"Genetics of gene expression and its effect on disease." Nature. (2008)
"Validation of candidate causal genes for obesity that affect..." Nat Genet. (2009)
….. Plus 10 additional papers in Genome Research, PLoS Genetics, PLoS Comp.Biology, etc
CVD "Identification of pathways for atherosclerosis." Circ Res. (2007)
"Mapping the genetic architecture of gene expression in human liver." PLoS Biol. (2008)
…… Plus 5 additional papers in Genome Res., Genomics, Mamm.Genome
Bone "Integrating genotypic and expression data …for bone traits…" Nat Genet. (2005)
d
..approach to identify candidate genes regulating BMD…" J Bone Miner Res. (2009)
Methods "An integrative genomics approach to infer causal associations ... Nat Genet. (2005)
"Increasing the power to detect causal associations… PLoS Comput Biol. (2007)
"Integrating large-scale functional genomic data ..." Nat Genet. (2008)
…… Plus 3 additional papers in PLoS Genet., BMC Genet.

28. List of Influential Papers in Network Modeling
 50 network papers
 http://sagebase.org/research/resources.php

29. (Eric Schadt)

30. “Data Intensive” Science- Fourth Scientific Paradigm
Score Card for Medical Sciences
Equipment capable of generating
massive amounts of data A-
IT Interoperability D
Open Information System D-
Host evolving computational models
in a “Compute Space F

31. We still consider much clinical research as if we we
hunter gathers - not sharing
.

32. TENURE
FEUDAL
STATES

33. Clinical/genomic data
are accessible but minimally usable
Little incentive to annotate and curate
data for other scientists to use

34. Mathematical
models of disease
are not built to be
reproduced or
versioned by others

35. Assumption that genetic alterations
in human conditions should be owned

36. Lack of standard forms for sharing data
and lack of forms for future rights and consentss

37. Publication Bias- Where can we find the (negative) clinical data?

38. sharing as an adoption of common standards..
Clinical Genomics Privacy IP

39. Sage Mission
Sage Bionetworks is a non-profit organization with a vision to
create a commons where integrative bionetworks are evolved by
contributor scientists with a shared vision to accelerate the
elimination of human disease
Building Disease Maps Data Repository
Commons Pilots Discovery Platform
Sagebase.org

40. Sage Bionetworks Collaborators
 Pharma Partners
 Merck, Pfizer, Takeda, Astra Zeneca,
Amgen, Johnson &Johnson
 Foundations
 Kauffman CHDI, Gates Foundation
 Government
 NIH, LSDF
 Academic
 Levy (Framingham)
 Rosengren (Lund)
 Krauss (CHORI)
 Federation
 Ideker, Califarno, Butte, Schadt 40

41. NEW MAPS
Disease Map and Tool Users-
( Scientists, Industry, Foundations, Regulators...)
PLATFORM
Sage Platform and Infrastructure Builders-
( Academic Biotech and Industry IT Partners...)
PILOTS= PROJECTS FOR COMMONS
Data Sharing Commons Pilots-
(Federation, CCSB, Inspire2Live....)
NEW TOOLS
ORM
APS
Data Tool and Disease Map Generators-
(Global coherent data sets, Cytoscape,
M
F
PLAT
Clinical Trialists, Industrial Trialists, CROs…)
NEW
RULES GOVERN RULES AND GOVERNANCE
Data Sharing Barrier Breakers-
(Patients Advocates, Governance
and Policy Makers, Funders...)

42. Bin Zhang
Integration of Multiple Networks for Jun Zhu
Pathway and Target Identification
CNV
Data
Gene
Expression
Clinical
Traits
Co-Expression
Bayesian Network
Network
Integration of Coexp. &
Bayesian Networks 42
Key Driver Analysis
42

43. Bin Zhang
Key Driver Analysis Jun Zhu
Justin Guinney
http://sagebase.org/research/tools.php 43

44. Bin Zhang
Model of Breast Cancer: Co-expression Xudong Dai
Jun Zhu
A) Miller 159 samples B) Christos 189 samples
NKI: N Engl J Med. 2002 Dec 19;347(25):1999.
Wang: Lancet. 2005 Feb 19-25;365(9460):671.
Miller: Breast Cancer Res. 2005;7(6):R953.
Christos: J Natl Cancer Inst. 2006 15;98(4):262.
C) NKI 295 samples
E) Super modules
Cell cycle
Pre-mRNA
ECM
D) Wang 286 samples Blood vessel
Immune
response
44
Zhang B et al., Towards a global picture of breast cancer (manuscript).

45. Bin Zhang
Model of Breast Cancer: Integration Xudong Dai
Jun Zhu
Conserved Super-modules
mRNA proc.
= predictive
Breast Cancer Bayesian Network
Chromatin
of survival
Extract gene:gene relationships for selected super-modules from BN and define Key Drivers
Pathways & Regulators
(Key drivers=yellow; key drivers validated in siRNA screen=green)
Cell Cycle (Blue) Chromatin Modification (Black) Pre-mRNA proc. (Brown) mRNA proc. (red)
45
Zhang B et al., Key Driver Analysis in Gene Networks (manuscript)

46. Developing predictive models of genotype specific sensitivity
to Perturbations- Margolin
Predictive model

47. Developing predictive models of genotype-specific
sensitivity to compound treatment
Gene8c
Feature
Matrix
Expression,
copy
number,
somaQc
mutaQons,
etc.
Predic8ve
Features
(biomarkers)
Cancer
samples
with
varying
degrees
of
response
to
therapy
Sensi8ve
Refractory
(e.g.
EC50)
47

48. 1
20
100
500
Feature
Features
Features
Features
Elastic net regression
48

49. Bootstrapping retains robust predictive features
49

50. Our approach identifies mutations in genes upstream of MEK
as top predictors of sensitivity to MEK inhibition
#3
Mut
NRAS
!"#$% &"#$% #1
Mut
BRAF
PD-­‐0325901
'"#(%
#312
Mut
NRAS
)*!+,-% #./0-11%
2/345-674+%
#9
Mut
BRAF
50
PD-­‐0325901

51. For 11/12 compounds, the #1 predictive feature in an unbiased
analysis corresponds to the known stratifier of sensitivity
#2
CML
lineage
CML lineage
#1
EGFR
mut
EGFR mut
#1
EGFR
mut
EGFR mut
#1
CML
lineage
#1
EGFR
mut
CML linage
EGFR mut
#1
ERBB2
expr
ERBB2 expr
How
accurate
would
predic8ve
#1
BRAF
mut
models
perform
for
diagnos8cs?
BRAF mut
#1
HGF
expr
HGF expr
#2
NRAS
mut
NRAS mut
BRAF mut
#1
BRAF
mut
#3
KRAS
mut
KRAS mut
#2
NRAS
mut
NRAS mut
BRAF mut
#1
BRAF
mut
#3
KRAS
mut
KRAS mut
#2
NRAS
mut
NRAS mut
BRAF mut
#1
BRAF
mut
#2
TP53
mut
TP53 mut
#3
CDKN2A
copy
CDKN2A copy
#1
MDM2
expr
MDM2 expr
51

54. Why not share clinical /genomic data and model building in the
ways currently used by the software industry
(power of tracking workflows and versioning

55. Leveraging Existing Technologies
Addama
Taverna
tranSMART

56. INTEROPERABILITY
SYNAPSE
Genome Pattern
CYTOSCAPE
tranSMART
I2B2
INTEROPERABILITY

57. sage bionetworks synapse project
Watch What I Do, Not What I Say Reduce, Reuse, Recycle
My Other Computer is Amazon
Most of the People You Need to Work with
Don’t Work with You

58. Six
Pilots
at
Sage
Bionetworks
CTCAP
Non-­‐Responders
Arch2POCM
ORM
S
MAP
The
FederaQon
F
PLAT
Portable
Legal
Consent
NEW
Sage
Congress
Project
RULES GOVERN

59. CTCAP
Clinical Trial Comparator Arm Partnership “CTCAP”
Strategic Opportunities For Regulatory Science
Leadership and Action
FDA
September 27, 2011

60. Clinical Trial Comparator Arm
Partnership (CTCAP)
 Description: Collate, Annotate, Curate and Host Clinical Trial Data
with Genomic Information from the Comparator Arms of Industry and
Foundation Sponsored Clinical Trials: Building a Site for Sharing
Data and Models to evolve better Disease Maps.
 Public-Private Partnership of leading pharmaceutical companies,
clinical trial groups and researchers.
 Neutral Conveners: Sage Bionetworks and Genetic Alliance
[nonprofits].
 Initiative to share existing trial data (molecular and clinical) from
non-proprietary comparator and placebo arms to create powerful
new tool for drug development.
Started Sept 2010

61. Shared clinical/genomic data sharing and analysis will
maximize clinical impact and enable discovery
• Graphic
of
curated
to
qced
to
models

62. Non-­‐Responders
Project
To identify Non-Responders to approved
Oncology drug regimens in order to improve
outcomes, spare patients unnecessary toxicities
from treatments that have no benefit to them, and
reduce healthcare costs

63. The
Non-­‐Responder
Cancer
Project
Leadership
Team
Stephen Friend, MD, PhD
Todd Golub, MD
Founding Director Cancer Biology
President and Co-Founder of
Program Broad Institute, Charles Dana
Sage Bionetworks, Head of
Investigator Dana-Farber Cancer
Merck Oncology 01-08,
Institute, Professor of Pediatrics Harvard
Founder of Rosetta
Medical School, Investigator, Howard
Inpharmatics 97-01, co-
Hughes Medical Institute
Founder of the Seattle Project
Richard Schilsky, MD
Garry Nolan, PhD Chief, Hematology- Oncology, Deputy
Professor, Baxter Laboratory of Stem Director, Comprehensive Cancer
Cell Biology, Department of Microbiology Center, University of Chicago; Chair,
and Immunology, Stanford University National Cancer Institute Board of
Director, Proteomics Center at Stanford Scientific Advisors; past-President
University
ASCO, past Chairman CALGB clinical
trials group
11

64. The
Non-­‐Responder
Project
is
an
internaQonal
iniQaQve
with
funding
for
6
iniQal
cancers
anQcipated
from
both
the
public
and
private
sectors
GEOGRAPHY
United
States
China
TARGET
CANCER
Ovarian
Renal
Breast
AML
Colon
Lung
RetrospecQve
FUNDING
SOURCE
Seeking
private
sector
and
study;
likely
to
Funded
by
the
Chinese
government
philanthropic
funding
for
be
funded
by
and
private
sector
partners
the
Federal
prospec8ve
studies
Government
5

65. Arch2POCM
Restructuring
the
PrecompeQQve
Space
for
Drug
Discovery
How
to
potenQally
De-­‐Risk
High-­‐Risk
TherapeuQc
Areas

67. The
FederaQon

68. How can we accelerate the pace of scientific discovery?
2008
2009
2010
2011
Ways to move beyond
“traditional” collaborations?
Intra-lab vs Inter-lab
Communication
Colrain/ Industrial PPPs Academic
Unions

69. (Nolan
and
Haussler)

70. human aging:
predicting bioage using whole blood methylation
• Independent training (n=170) and validation (n=123) Caucasian cohorts
• 450k Illumina methylation array
• Exom sequencing
• Clinical phenotypes: Type II diabetes, BMI, gender…
Training Cohort: San Diego (n=170) Validation Cohort: Utah (n=123)
RMSE=3.35 RMSE=5.44
100
100
! ! !!
!
! !
!! ! !
! !
! !
! ! ! ! !! !
! ! ! !!!!
!! ! !
! ! ! !! !
! !! !
! !!
! ! !! !
80
Biological Age
Biological Age
!! !
!!!! !!
! !! !
! ! ! !!! !
80
! !!!
!! ! ! ! !
!!
! !
! !
! !! !
! ! ! !!!!
!!! !
!!!
!!
! !! !!!!
! !!! !! !! ! ! !
! ! ! !
! !
! !!! !
! !! !
! !!
! !!
!! ! !! !
!! !
! !!!! ! !
! ! !! !
!! !
! ! !! !
! !!!
! ! ! !!! ! !!
!!
!! !
!! ! !
! !
!! ! !
!
! !!
!! ! ! !
!
60
!!! ! ! !
! !! !
! ! !
! !! !
60
! ! ! !!
!! ! ! !
! !
!!
! !!!!
! ! !
! ! !! !
! !!
! ! ! !!
! ! !
! ! ! 40
40
! !
40 50 60 70 80 90 100 40 50 60 70 80 90
Chronological Age Chronological Age

71. sage federation:
model of biological age
Faster Aging
Predicted
Age
(liver
expression)
Slower Aging
Clinical Association
- Gender
- BMI
- Disease
Age Differential Genotype Association
Gene Pathway Expression
Chronological
Age
(years)

72. Reproducible
science==shareable
science
Sweave: combines programmatic analysis with narrative
Dynamic generation of statistical reports
using literate data analysis
Sweave.Friedrich Leisch. Sweave: Dynamic generation of statistical reports
using literate data analysis. In Wolfgang Härdle and Bernd Rönz,editors, Compstat 2002 –
Proceedings in Computational Statistics,pages 575-580.
Physica Verlag, Heidelberg, 2002. ISBN 3-7908-1517-9

73. Federated
Aging
Project
:
Combining
analysis
+
narraQve
=Sweave Vignette
Sage Lab
R code + PDF(plots + text + code snippets)
narrative
HTML
Data objects
Califano Lab Ideker Lab Submitted
Paper
Shared
Data
JIRA:
Source
code
repository
&
wiki
Repository

74. Portable
Legal
Consent
(AcQvaQng
PaQents)
John
Wilbanks

78. Sage
Congress
Project
April
20
2012
RA
Parkinson’s
Asthma
(Responders
CompeQQons)

79. Why not use data intensive science
to build models of disease
Current Reward Structures
Organizational Structures and Tools
Six Pilots
Opportunities

80. Actionable Cancer Network Models
And Open Medical Information Systems
IMPACT
ON
PATIENTS