This document summarizes network medicine and its applications. It discusses how human diseases can be modeled and studied as complex networks. Disease genes are found to cluster together in protein interaction networks, forming disease modules. Mapping disease genes onto interactome networks can help identify new candidate genes and delineate disease modules. Validation using various biological data shows the predicted disease genes are statistically associated with the disease. Mapping asthma genes in this way identified a statistically significant disease module within the first 200 prioritized genes. Network medicine approaches provide a framework for understanding the molecular basis of diseases.

1. Network Medicine
Albert-László Barabási
Center for Complex Networks Research
Northeastern University
Department of Medicine and CCSB
Harvard Medical School
Central European University, Budapest
www.BarabasiLab.com

3. DISEASOME PHENOME
GENOME
Gene Disease
network network
Goh, Cusick, Valle, Childs, Vidal & Barabási, PNAS (2007)

4. Human Disease Network

5. Human Protein-Protein Interaction Network
Rual et al. Nature 2005; Stelze et al. Cell 2005

6. World Wide Web
Exponential Network Nodes: WWW documents
Links: URL links
Over 10 billion documents
ROBOT: collects all URL’s
found in a document and
follows them recursively
Scale-free Network
P(k) ~ k-
R. Albert, H. Jeong, A-L Barabási, Nature, 401 130 (1999).

8. Metabolic Network Protein Interactions
Jeong, Tombor, Albert, Oltvai, & Barabási, Nature (2000); Jeong, Mason, Barabási &.
Oltvai, Nature (2001); Wagner & Fell, Proc. R. Soc. B (2001)

9. Robustness of scale-free networks
node
failure
Failures Attacks
Albert, Jeong, Barabási, Nature 406 378 (2000)

10. Hubs and Essentiality
18% 24% 62%
 Hubs evolve slower: they are more alike in different organisms
[H Fraser et al., Science (2002). Krylov, et al. Genome Res.(2003)]
 Hub removal has more phenotypic consequences [Yu et al. Science (2008)]
Jeong, Mason, Barabási, and Oltvai, Nature 411, 41-42 (2001

11. Are Disease Genes Hubs?
Goh, Cusick, Valle, Childs, Vidal & Barabási, PNAS (2007)

13. Essential Proteins are Hubs;
Disease Genes are at the Network Periphery
Goh, Cusick, Valle, Childs, Vidal & Barabási, PNAS (2007)

14. Human Disease Network

16. ~ 13’039’018 patients
~ 32’341’348 records
(hospitalizations)
Identifier, Time of Visit, State, Age, Gender, Poverty (0-1),
Up to 10 diagnosis (from ~950 cat (3digit) / ~14000 (5digit) )

17. I1 I2
N
Disease 1 C12 Disease 2
Cij N Cij N  I i I j
 ij  ij 
Ii I j I i I j ( N  I i )( N  I j )
Park, Lee, Christakis, Barabási, Mol Syst Biol (2008)

18. Genetic Associations and Comorbidity Measures
(RR, or Relative Risk)

19. Average Comorbidity Values
Park, Lee, Christakis, Barabási, Mol Syst Biol (2008)

20. Phenotypic Disease Network (PDN)
I1 I2
N
Disease 1 C12 Disease 2
Cij N Cij N  I i I j
 ij  ij 
Ii I j I i I j ( N  I i )( N  I j )
(RR or Relative Risk)

Hidalgo, Blumm, Barabasi & Christakis, PLOS Comp. Biol. (2009)

21. http://hudine.neu.edu/
Hidalgo, Bloom, A-L. B., Christakis, PLOS Comp. Biol. (2009); A. Rzhetsky , PNAS (2007)

22. Network Position and Survival Rate

Hidalgo, Blumm, Barabasi & Christakis, PLOS Comp. Biol. (2009)

23. CONTROLLABILITY
A system is controllable if it can be driven from
any initial state to any desired final state in
finite time.
R. E. Kalman, J.S.I.A.M. Control (19

24. FuturICT: Taming Complexity
BarabasiLab.com

25. CONTROLLABILITY: What did we learn?
Organizational Network: 1-10%
Regulatory Network: 80%
• Driver nodes avoid the hubs.
• The more interconnected a network (high <k>), the fewer driver nodes we need.
• The more uniform the node degrees, the fewer driver nodes we need.
• Sparse and heterogeneous networks are hardest to control (i.e. most real
networks).
Y.-Y. Liu, J.-J. Slotine, A.-L. Barabási, Nature (20

27. Zoom

28. Calling Patterns at Midnight and Noon
Busy@Midnight Sleep@Noon
Sleep@Midnight Busy@Noon
Midnight Noon

29. Human Protein-Protein Interaction Network
Rual et al. Nature 2005; Stelze et al. Cell 2005

30. Local clustering of disease genes: disease modules
Cellular components that form a topological module have have closely related function, thus they correspond to a
function module, and a disease is a result of the breakdown of functional module
Human Interactome
Asthma
Nodes color correspond to different diseases from
OMIM/GWAS studies

31. Local clustering of disease genes: disease modules
Cellular components that form a topological module have have closely related function, thus they correspond to a
function module, and a disease is a result of the breakdown of functional module
Human Interactome
Asthma
Nodes color correspond to different diseases from
OMIM/GWAS studies

32. Genes that are involved in the same disease show
a high propensity to interact with each other
Each axis represent the category of disease associated
with the protein in an interaction pair
Gandhi et al Nat Genet.
2006, :285-93
Goh, Cusick, Valle, Childs, Vidal & Barabási, PNAS (2007)
OMIM disease genes and clustering

33. Mapping out disease modules
Barabási, Gulbahce, Loscalzo (2010)

34. Interactome Reconstruction
Data Links Nodes Experimental/Predictive
Metabolic
5287
Regulatory
Binary Interaction
Experimental/(Binary
Network (binary 1280
considered from Intact and
interactions from
Mint database)
Y2Hybrid, Intact and Mint) Literature curated
Kinase
42420
15315 6101 CORUM
1620 4197
2936
Binary
Binary Interaction 27837 172
7190 3302 Experimental/Binary 12775
network Y2Hhybrid only 403
1929
Transfac regulatory
1340 781 Experimental
network
Metabolic coupling
Interaction network 10642 921 Experimental/ Predictive
(BIGG/KEGG)
Curated HPRD-Biogrid-
74195 10890 Literature curated
Intact-Mint
CORUM-All complexes
31276 2069 Experimental
data
Kinase-substrate pairs 327 (kinases)
6110 Experimental/Literature
(PhosphoSitePlus,
1771 (substrates)
Phospho.ELM)

35. Seed Genes: Disease genes associated with asthma
Sources Genes References (PubMed id)
FCER1B,HLA-DRB1,HLA-DQB1,HLA-
DPB1,STAT6,NR3C1,GFRA2,GATA2,SH2B3,IKZF2,GST
Literature, as compiled in Vercelli et al M1,GSTP1,GSTT1,FLG,IL10,IL18,CTLA4,HAVCR1,GPR
2008 A,NAT2,NOS1,CMA1,ACE,TBXA2R,CTA,SCGB1A1 18301422
CCL17,CCL22,CCL2,CD200,CX3CR1,CXCR1,CXCR2,IL8,
Chemokines CCL11,
CCL11TNFRSF4TNFSF4ICOSC5CHRNA3TRPA1TRPV1H
Chemokines RH4
Co-stim pathway TNFRSF4,TNFSF4,ICOS
Complement pathway C5
GPCR-ligand inflammatory mediators CHRNA3,TRPA1,TRPV1,HRH4,PTGDS,GPR44,HNMT,P
pathway TGER2
Growth factors pathway F2RL1,FGF2,PDGFB
CSF2,IKBKB,NGF,PDE4,PIK3CA,STAT1,MAPK14,MAPK
Inflammatory signaling pathway 8,MMP12,SYK,IRAK3,OSM,TNFA
ALOX5AP,LTB4R2,LTA4H,LTC4S,ALOX5,IFNB1,TLR7,T
Leukotriene pathway LR3,TLR9,TLR4
Pathogen response pathway IFNB1,TLR7,TLR3,TLR9,TLR4
Proteases pathway MMP9,TPSAB1,SPINK5,ELANE
Structural Genes/Mucosal Epithelial
Altogether: pathway CTNNA3,MUC7,TGFB1,CLCA1,MUC5AC
Th17 pathway IL17,IL23A,IL6
145 genes Th2 pathway IL13RA2,IL4,IL33,IL4R,IL9,IL25
ADORA2A,SRC,CD28,IL21,IL22,EGF,CD40
EDN1,GSDMA,GSDMB,HLA- 21103062,20860503,20622879,2015
DQA1,IL18R1,IL2RB,MAVS,MYB,PDE11A,PDE4D,RAD 9242,19198610,20920776,19426955
50,RORA,SCG3,SLC22A5,SMAD3,VDR,IL13,CD14,DPP ,20159242,20159242,20881960,202
GWAS-Asthma 10,ORMDL3,IL12B,PTGDR 08534,17611496
GWAS-Childhood Asthma CRB1,DENND1B,CHCHD9,TLE4,ADAM33,CCL5 20032318,19714205,21103062
GWAS-Plasma eosinophil count IL1RL1,IL5,TSLP 19198610
GWAS-YKL-40 levels CHI3L1 18403759
ADRB2,ALOX5,IGHE,HNMT,MUC7,PHF11,SCGB3A2,T
BX21,SCGB1A1,CCL11TNFA,LTA4H,LTC4S,IL4,IL33,IL4
R,IL9,TLR9,SPINK5,CTNNA3,CRHR1,IL8TLR4CLCA1,M
MeSH term associated with asthma UC5AC,IL25
CCR3,GPR44,PTGER2,IRAK3,ASRT3,ASRT4ASRT6HLA-
G,PLA2G7ADRB2IGHEALOX5,HNMT,MUC7,PHF11,SC
OMIM genes realeated to asthma GB3A2,TBX21,SCGB1A1,CCL11,TNFA

36. Mapping out disease modules
Barabási, Gulbahce, Loscalzo (2010)

37. Seed genes clusters in Human Interactome
Of the 145 seed genes:
126 on the interactome.
37: giant component.
77: isolated

38. Disease Module Detection

39. Mapping out disease modules
Barabási, Gulbahce, Loscalzo (2011)

40. Biological data for validation
Differentially
expressed gene
set Gene
GSE3183 (human
airway Ontologies(biological
hyperresponsiveness pathway)/ MSIgDB
GSE473 (Human CD4+ MSIgDB-Molecular signature
lymphocytes) database: GeneGo pathways for
GSE470 (Human asthma 6700 gene sets containing Asthma
exacerbatory data from
factors)GSE3004 KEGG, Biocarta, canonical
(Human airway and reactome pathways 35 pathways with 737 genes Genes associated with
to be significantly enriched
epithelial) Diseases comorbid
GSE2125 (alveolar for asthma
with Asthma (493)
macrophages; p<0.02)
Genes associated with
and human primary cell
lines of NHBE, NHLF and Superarray diseases comorbid to asthma
with relative risk relative risk
BSMC, exposed to pathways (RR) score >1.5 –Medicare
cytokines or PBS 38 pathways specific to data
(p<0.05) asthma Genes associated with
diseases comorbid to
asthma with relative risk
relative risk (RR) score >2
–I3 data

41. Validation of the predicted disease genes
Do the predicted genes show a statistically
significant biological association with asthma?
Can we define the disease module boundary?

42. Biological validation
of prioritized DMD genes
 Over all statistical significance
appear to be limited to roughly
to the first 200 genes selected
by the method.
 Beyond these genes the
statistical significance gradually
vanishes, indicating that the
genes later added may not be
part of the disease module

43. Final Candidate Genes Prioritization

44. GeneGo pathways and 200 prioritized candidate genes
Community 1 enriched in most of pathways

45. Network Medicine
Barabási, New England Journal of Medicine (2007)