P System Model Op/misa/on by Means of Evolu/onary Based Search Algorithms C. García‐Mar+nez, C. Lima, J. Twycross, M. Lozano, N. Krasnogor 1
May 11, 2015
PSystemModelOp/misa/onbyMeansofEvolu/onaryBasedSearch
Algorithms
C.García‐Mar+nez,C.Lima,
J.Twycross,M.Lozano,N.Krasnogor
1
Outline
• Mo+va+on:Systems&Synthe+cBiology,PSystemsbasedmodeling
• Methods&ExperimentalSetup
• ResultsandDiscussion
• Conclusions
2
/136
•The Cell senses the environment and its own internal states•Makes Plans, Takes Decisions and Act•Evolution is the master programmer
3
The Cell as an Intelligent (Evolved) Machine
Cell
Internal States
Environmental Inputs
Actions
Amir Mitchell, et al., Adaptive prediction of environmental changes by microorganisms. Nature June 2009.
3
/136
•The Cell senses the environment and its own internal states•Makes Plans, Takes Decisions and Act•Evolution is the master programmer
3
The Cell as an Intelligent (Evolved) Machine
Cell
Internal States
Environmental Inputs
Actions
Amir Mitchell, et al., Adaptive prediction of environmental changes by microorganisms. Nature June 2009.
Wikimedia Commons
3
/136
Network Motifs: Evolution’s Preferred Circuits•Biological networks are complex and vast•To understand their functionality in a scalable way one must choose the correct abstraction
•Moreover, these patterns are organised in non-trivial/non-random hierarchies
•Each network motif carries out a specific information-processing function
4
“Patterns that occur in the real network significantly more often than in randomized networks are called network motifs” Shai S. Shen-Orr et al., Network motifs in the transcriptional regulation
network of Escherichia coli. Nature Genetics 31, 64 - 68 (2002)
Radu Dobrin et al., Aggregation of topological motifs in the Escherichia coli transcriptional regulatory network. BMC Bioinformatics. 2004; 5: 10.
4
/1365
Y positively regulates X
Negative autoregulation
Positive autoregulation
The C1-FFL is a ‘sign-sensitive delay’ element and a persistence detector.The I1-FFL is a pulse generator and response accelerator
U. Alon. Network motifs: theory and experimental approaches. Nature Reviews Genetics (2007) vol. 8 (6) pp. 450-461
Shai S. Shen-Orr et al., Network motifs in the transcriptional regulation network of Escherichia coli. Nature Genetics 31, 64 - 68 (2002)
5
/1366
Shai S. Shen-Orr et al., Network motifs in the transcriptional regulation network of Escherichia coli. Nature Genetics 31, 64 - 68 (2002)
•The correct abstract ions facilitates understanding in complex systems.
•Provide a route to engineering , programming and evolving cells and their models
6
7
7
• Cells(andmostbiologists)don’tdodifferen/alcalculus!
• Psystemsareaexecutablespecifica/onsthatcloselymimicbiologicalreality.
• Theseareprogramsthatexplicitlymimictheinternalbehaviorofcellsystems.
• Theseprogramsareexecutedinavirtualmachinethatcapturestheintrinsicstochas5cityinherentinbiology
7
FundamentalECChallenge
• Learningaprogramwithstochas/cbehaviorvs.learningaPsystem.
•A cell is a living example of distributed stochastic computing.
function f1(p1,p2,p3,p4){if (p1<p2) and (rand<0.5) print p3else print p4}
function f1(p1,p2,p3,p4){if (p1<p2) RND print p3 RNDelse RND print p4 RND}
8
ModularAssemblyofPSystems
• Modules:setofrulesrepresen/ngmolecularinterac/onsthatoccuroNen.
• Elementalmodules:Degrada/on,complexa/on,unregulatedgeneexpression,nega/vegeneexpression,etc.
• Combinatorics:Combina/onofbasicmodules(building‐blocks)originatesmorecomplexmodules,allowingmodularandhierarchicalmodellingwithPsystems.
• Challenge:Explorethelargecombinatorialspaceofmodulesandcorrespondingparameters.
9
/136
Multi-Objective Optimisation in Morphogenesis
Rui Dilão, Daniele Muraro, Miguel Nicolau, Marc Schoenauer. Validation of a morphogenesis model of Drosophila early development by a multi-objective evolutionary optimization algorithm. Proc. 7th European Conference on Evolutionary Computation, ML and Data Mining in BioInformatics
(EvoBIO'09), April 2009.
1010
/136
Parameter Optimisation in Metabolic Models
11
A. Drager et al. (2009). Modeling metabolic networks in C. glutamicum: a comparison of rate laws in combination with various parameter optimization strategies. BMC Systems Biol ogy 2009, 3:5
11
/136
Evolving P Systems Structures
12
F. Romero-Campero, H.Cao, M. Camara, and N. Krasnogor. Structure and parameter estimation for cell systems biology models. Proceedings of the Genetic and Evolutionary Computation Conference (GECCO-2008), pages 331-338. ACM Publisher, 2008.
H. Cao, F.J. Romero-Campero, S. Heeb, M. Camara, and N. Krasnogor. Evolving cell models for systems and synthetic biology. Systems and Synthetic Biology , 2009
12
Outline
• Mo/va/on:Systems&Synthe/cBiology,PSystemsbasedmodeling
• Methods&ExperimentalSetup
• ResultsandDiscussion
• Conclusions
13
Methods&ExperimentalSetup
• Comparedifferentevolu/onaryalgorithmstoop/miseparameters(kine/cconstants)inPsystems.
• Fourtestcasesofincreasingdifficultyanddimension:1.TC1:Pulsegeneratorfordifferentini/alcondi/ons(13parameters).
2.TC2:SameproblemasTC1butwithalargerparameters’domain.
3.TC3:Moregeneralpulsegenerator:feed‐forwardloopmo/f(18parameters).
4.TC4:Bandwidthdetector(34parameters).
• Experimentalbudgetwasrestrictedto1000func/onevalua/ons.
14
TargetModels
15
TargetModels
•HighlyDimensional•Noisy&Uncertainoutcomes•Non‐lineari+es•ExpensiveFunc+onevalua+ons
15
TargetModels
•HighlyDimensional•Noisy&Uncertainoutcomes•Non‐lineari+es•ExpensiveFunc+onevalua+ons
Op+misa+onHell!
15
Evolu/onaryAlgorithms
• CovarianceMatrixAdapta/onAlgorithm(CMA‐ES)
• Differen/alEvolu/on(DE)
• Opposi/on‐BasedDifferen/alEvolu/on(ODE)
• Real‐CodedGene/cAlgorithm(GA)
• VariableNeighbourhoodSearchwithEvolu/onaryComponents(VNS‐ECsv1andv2)
16
ExperimentalDetails
• Fitnessofagivencandidatesolu/onisgivenby:1. RunthecorrespondingPsystemwiththe
mul/compartmentGillespiestochas/csimula/onalgorithm(20runs).
2. Averagetheoutput/meseriesofallrunsandcalculatethedifferencetothetargetseries,usingtherandomlyweightedsummethod.
• Be\er(forguidingthesearch)thansimpleconsideringtheRMSE,par/cularlywhen/meseriesrangeroverdifferentscales.
• Op/misa/onresultsareaveragedover50runs.
17
Outline
• Mo/va/on:Systems&Synthe/cBiology,PSystemsbasedmodeling
• ExperimentalSetup
• ResultsandDiscussion
• Conclusions
18
Results
19
Discussion
• AlgorithmsrankedaccordingtoRMSE.• Mann‐WhitneyUtestwithp‐value=0.05todeterminewhichalgorithmsperformsignificantlybe\erthanothers.
• ForTC1,mostalgorithmsperformequallywell,withexcep/ontoCMA‐ESandVNS‐EC1.
• ForTC2,wecanfindsignificantdifferencesbetweenalgorithms,whereGAisthebe\er.
• Reducingbiologicalknowledge(fromTC1toTC2)clearlyaffectstheperformanceofthealgorithms.
20
Discussion
• ForTC3,manyalgorithmsperformsimilar,butDE,ODE,andGAseemtoperformslightlybe\er.SimilartoTC1butnowVNS‐EC2performsconsiderablyworse.
• ForTC4,wherethereisalargernumberofparameters,resultsaresignificantlydifferentfromotherproblems.
• VNS‐ECsnowperformsignificantlybe\erthanremainingapproaches.
• CMA‐ES,ODE,andDEperformsimilarly,whileGAistheleastcompe//ve.
21
Discussion
• Ingeneral,GA,ODE,andDEperformbe\erforproblemswithfewparameters(13and18).GAperformsbe\erwhenbiologicalknowledgeisreduced.
• Ontheotherhand,VNS‐ECsperformbe\erforthelargerproblem(38parameters).
• Whyisthis?Thenumberofevalua/onsallowedissmall(1000).Let’shavealook…
22
BestFitness
23
BestFitness
• Twoimportantobserva/ons:1. ForTC4(largernumberofparameters),GA,DE,and
ODEreducetheirconvergencespeedsbecauseevolvingpopula/onsofindividualsconsumesmanyresources.However,VNS‐ECswhichfocusthesearchononesolu/onmakeabe\erusageofthereducedbudget.
2. Whentheprobleminvolvesfewerparameters,theallowedbudgetisenoughtoproperlyconvergeapopula/onofsolu/ons.Inthiscase,VNS‐ECsarenotcompe//veanymore.
24
AverageModelFit
• TestCase1
• TestCase2
25
AverageModelFit
• TestCase3
Forprotein1,allalgorithmshavesimilaroutputtothetarget.
26
AverageModelFit
• TestCase4
27
Outline
• Mo/va/on:Systems&Synthe/cBiology,PSystemsbasedmodeling
• Methods&ExperimentalSetup
• ResultsandDiscussion
• Conclusions
28
Conclusions• Considered4testcasesofincreasingdifficulty.• Limitedcomputa/onalresources(1000evalua/ons)havebeenimposedgiventheincreased/metoevaluatecandidatesolu/ons.
• Forthisexperimentalsetup,ithasbeenfoundthat:1. Whennumberofkine/cconstantsissmall,GA,DE,andODEare
robustop/misers.2. Whennumberofparametersincreases,theVNS‐Ecsobtain
be\erresults.
• DE(and,notreported,PSO)giveagoodcompromiseofquality/speedandconfigura/oneffortforsmalltomediumsizeproblems.
• Forlargerproblems,VNS‐Ecs(withouttoomanyparameters)seemthewaytogo.
• Fitnesscriterionmustberevisited!!!
29
/136
Acknowledgements
•Jonathan Blake
•Claudio Lima
•Francisco Romero-Campero
•Karima Righetti
•Jamie Twycross
Integrated Environment
Machine Learning & Optimisation
Modeling & Model Checking
Molecular Micro-Biology
Stochastic Simulations
Members of my team working on SB2
EP/E017215/1
EP/H024905/1
BB/F01855X/1
BB/D019613/1
University of NottinghamProf. M. Camara, Dr. S. Heeb, Dr. G. Rampioni, Prof. P. WilliamsWeizmann Institute of ScienceProf. D. Lancet, Prof. I. Pilpel
3030
31
31
31
31