prof. csermely péter semmelweis egyetem, link-csoport

Prof. Prof. Csermely PéterCsermely PéterSemmelweis Egyetem, LINK-csoportSemmelweis Egyetem, LINK-csoport

Nagyméretű, ritka gráfok moduláris, játékelméleti és perturbációdinamikai

elemzése biológiai és más valós hálózati

problémák megoldásában

[email protected]

Advantages of network multi-disciplinarity

Watts & Strogatz,1998

Networks have general properties• small-worldness• hubs (scale-free degree distribution)• nested hierarchy• stabilization by weak links

Karinthy,1929

Generality of network properties offers• judgment of importance• innovation-transfer across different layers of complexity

Barabasi & Albert, 1999

Csermely, 2004; 2009

Example to break conceptual barriers

ecosystem, market, climate• slower recovery from perturbations• increased self-similarity of behavior• increased variance of fluctuation-patternsNature 461:53

Aging is an early warning signal of a critical transition: deathPrevention: central elements with less predictable behavior• omnivores, top-predators• market gurus • stem cells

Farkas et al, Science Signaling 4:pt3

Less predictable (creative) elementshave a large dynamic centrality

Problem solver: specialized to a task, predictablechange of roles

Creative: few links to hubs, unexpected re-routing, flexible, unpredictable

Distributor: hub, specialized to signal distribution, predictable

Csermely, Nature 454:5TiBS 33:569TiBS 35:539

Creative amino acids• centre of residue-network• in structural holes

Cyt-P450(CYP2B4)

drug-binding oxidation

Creative proteins• stress proteins• signaling switches

Creative cells• stem cells• our brain

Creative persons• firms• societies

Creative elements are central and…Creative elements are central and…

mobile

Csermely, Nature 454:5TiBS 33:569TiBS 35:539

Robert: broker – netokratJames: looser

Creative elements of social networks

Ron Burt, Structural holes, Harvard Univ. Press 1992

3 examples of network modelling

• topological centrality is central in network perturbations(Turbine: www.linkgroup.hu/Turbine.php)

• community centrality: prediction of survival importance(ModuLand: www.linkgroup.hu/modules.php)

• game-centrality: prediction of biological regulators(NetworGame: www.linkgroup.hu/NetworGame.php)

Turbine algorithm assessing Turbine algorithm assessing network perturbation dynamicsnetwork perturbation dynamics

any real networks can be added, modifiednormalizes the input networkany perturbation types (multiple, repeated, etc.)any models of dissipation, teaching and agingMatlab compatible

www.linkgroup.hu/Turbine.phpFarkas et al, Science Signaling 4:pt3

Hubs + inter-modular nodes are primary transmitters of network perturbations

www.linkgroup.hu/Turbine.php

Start: module-center hub Starting node: bridge

Farkas et al., Science Signaling 4:pt3

3 examples of dynamic centrality




networks:• STRING 7 (5329/190018)• BioGRID (5329/91749)• Ekman (2444/6271)

mRNA expressionrestingprotein weight

resting link-weight

multiplication

weights:• equal units• proteins (Nature 441:840)• mRNA-s (Cell 95:717)

13 stress conditions, 65 experiments(Gasch et al, Mol. Biol. Cell 11:4241)

averagestress

discreteweights

continuousweightsaverage

uniquestress

6 stress conditions, 32 experiments(Causton et al, Mol. Biol. Cell 12:323)

stressedprotein weight

stressed link-weight

multiplicationaverage

Importance of modular overlapsImportance of modular overlapsof protein-protein interaction networks in stress

analysis of overlapping network modules:ModuLand method

Mihalik & CsermelyPLoS Comput. Biol. 7:e1002187

ModuLand method family detectsovelapping network communities

communityheaps of allnodes/links

communitylandscape

networkhierachy

communitiesas landscape hills

Kovacs et al, PLoS ONE 5:e12528www.linkgroup.hu/modules.phpnetwork of network scientists; Newman PRE 74:036104

Community heaps: NodeLand methodCommunity heaps: NodeLand method

starting node

community-44: 1127 schoolchildren, 5096 friendships; Add-Health

influence zones

communityheaps of allnodes/links

networkhierachy

communitiesas landscape hills

Kovacs et al, PLoS ONE 5:e12528www.linkgroup.hu/modules.phpnetwork of network scientists; Newman PRE 74:036104

available as a Cytoscape plug-in

communitycentrality:a measureof the influenceof all other nodes

communitylandscape

ModuLand method family detectsovelapping network communities

Mihalik & CsermelyPLoS Comput. Biol. 7:e1002187

Changing community centralitiesreflect the importance of survival in stress

proteinsynthesis

energydistribution

energydistribution

proteindegradation

survivalprocesses

• yeast protein-protein interaction network: 5223 nodes, 44314 links• stress: 15 min 37°C heat shock• link-weight changes: mRNA expression level changes

Mihalik & Csermely PLoS Comput. Biol. 7:e1002187

Changes of yeast interactome in crisis:a model of systems level adaptation

• BioGrid yeast interactome: 5223 nodes, 44314 links• stress: 15 min 37°C heat shock + Gasch et al. MBC 11:4241• link-weight changes: mRNA expression level changes

Stressed yeast cell:• nodes belong to less modules• modules have less intensive contactssmaller overlaps between modules

1. Stress-induced overlap decrease is general• proteins: water, stretch• brain: disease states, such as Alzheimer• animal compementary coop.: alpha-male• social dimensions: dissociate in stress, firms• ecosystems: patch separation in desiccation

Derenyi et al. Physica A 334:583 Csermely: Weak Links (Springer 2009)

stress

apoptosisassembly >

disassemblydisassembly >

assembly

network diameterdegrees of

freedom

2. Topological phase transitions reflect the overlap-decrease at one level higher

Consequences of network crisisConsequences of network crisis

Szalay et al, FEBS Lett. 581:3675; Palotai et al. IUBMB Life 60:10Mihalik & Csermely. PLoS Comput. Biol. 7:e1002187

stress increased network flexibility• spared links• noise and damage localization• modular independence: larger response-space and better conflict management

celldeath

networkdesintegrationcreative

elements*

Crisis survival: creative elements

*Schumpeterian creative destruction

Szalay-Bekő et al. in preparation

Bridges of Met-tRNA-synthase network identify signaling amino acids

Ghosh et al. PNAS 104:15711

signaling amino acidsother amino acids

Overlaps of signaling pathways are most pronounced in humans

growing prevalence of signaling cross-talks

Korcsmaros et al,Bioinformatics 26:2042;PLoS ONE 8:e19240www.SignaLink.org – 2010 version:646 proteins, 991 links in humans

3 examples of dynamic centrality




NetworGame program

• any model or real networks can be added (weighted, directed)• any game types (prisoners’ dilemma, hawk-dove, stag hunt, etc.)• any strategy update rules• synchronized, asynchron, semi-synchron update• starting strategies can be set individually by elementsFarkas et al., Science Signaling 4:pt3www.linkgroup.hu/NetworGame.php

Q-learning helps cooperation

replacing37 linksfrom 900

PD-gamebest-takes

over

18% cooperators3.5% cooperators

replacing37 linksfrom 900

PD-gameQ-learning

18% cooperators18% cooperators

Wang et al. PLoS ONE 3:e1917

regularnetwork

smallworldnetwork

Q-learning 1. helps cooperation even at high temptation levels2. stabilizes cooperation at different network topologies

SAME with• other short term strategy update rules• other network topologies• other games (extended-PD or Hawk-Dove)

1

2

Wang, Szalay, Zhang & CsermelyPLoS ONE 3:e1917

Long-term learning1. helps2. BUT does not stabilize cooperation


12


Long-term learning + innovation1. helps cooperation even at high temptation levels2. stabilizes cooperation at different network topologies


12


SAME with• other short term strategy update rules• other network topologies• other games (extended-PD or hawk-dove)

Learning + innovationexpand cooperative network topologies

imitation(best-takes-over)

learningQ-learninglearning + innovation


we are do not depend(that much…)on the network around us

Game-centrality: prediction of key nodes of social regulation I.hispanic old

young

BCBC

BCunion leaders: strikesociogram leaders: work

Farkas et al., Science Signaling 4:pt3www.linkgroup.hu/NetworGame.phpMichael’s strike network; Michael, Forest Prod. J. 47:41

Hawk-dove game (PD game: same)Start: all-cooperation = strikeStrike-breaker: defectsBC-s are the best strike-breakers

hawk-dove game:50-50% random starting D/Crecovers network modules

doves

hawks

instructor

administrator

BCBC

BCBC

BC

prisoners’ dilemma/stag hunt gamesmost influential single defective elementslargest betweenness centrality

Farkas et al., Science Signaling 4:pt3www.linkgroup.hu/NetworGame.phpZachary karate club network; J.Anthropol. Res. 33:452

Game-centrality: prediction of key nodes of social regulation II.

Game-centrality: prediction of key nodes of biological regulation

Farkas et al., Science Signaling 4:pt3www.linkgroup.hu/NetworGame.php

• Met-tRNA-synthase protein structure network signaling amino acids (Ghosh, PNAS 104:15711) largest game-centrality• yeast protein-protein interaction network amino acids regulating evolution (Levy, PLoS Biol 6:e264): large game centrality

www.csermelyblog.hu

Bemutatkozás

A könyvek innen tölthetőek le:

István A. Kovács

Máté Szalay-Bekő

ÁgostonMihalik

Stress

ModuLand

Eszter Hazai

Robin Palotai

Kristóf Z. Szalay

maybe You as acollaborator

Acknowledgments

Gábor I. SimkóTurbine:

perturbation NetworGame

Aging

Shijun Wang

[email protected]

half from the BME…

Turbine algorithm: perturbation model

dissipation of perturbations

‘learning’ – ‘aging’: changes of link weights attenuation of other links, if a link gains weight

Es, free energy of starting node; Ee, free energy of the other node on the link; l, degree; w, link weight; D0, dissipation constant; Cs, amplification constant; A, aging sensitivity;Cw, attenuation constantif perturbation exceeds a limit: links are exchanged to half as many random links

www.linkgroup.hu/Turbine.phpFarkas et al., Science Signaling 4:pt3

Game rules• canonical Prisoner’s Dilemma game: CC:R(3); CD:S/T(0/3-6); DD:P(1)

extended PD game: CC:R(1); CD:S/T(0/1-2); DD:P(0) Hawk-Dove game: CC:G/2 (C>G); CD:G; DD:-(G-C)/2

• 2,500 agents are tied to network nodes (no evolution)• agents can play with their neighbors only• they can either defect or cooperate• 50-50% defectors and cooperators start

at a random distribution• 5,000 plays, % of cooperators on last 10 rounds• average of 100 games is displayed

Wang, Szalay, Zhang and CsermelyPLoS ONE 3:e1917

Strategy update rules• short-term strategy update: best takes over (imitation, copy the richest neighbor)

replicator dynamics (pair-wise comparision dynamics; copy randomly selectedneighbor, if richer); proportional updating (spread to all neighbors, if richer)

• long-term learning, innovative strategy update: Q-learning (a type of reinforcement learningoptimal strategy to maximize total discounted expected reward; annealing temp 100, discount factor 0.5)

• long-term learning: accumulated payoff not only last round

• innovation: opposite of agent’s strategy with p (0.0001) probability

• synchronized update, averaged payoffs

Wang, Szalay, Zhang and CsermelyPLoS ONE 3:e1917

prof. csermely péter semmelweis egyetem, link-csoport

Documents