tm Series FROM NEURON TO NETWORK MEASUREMENT, ANAL

tmSeriesFROMNEURONTONETWORK:MEASUREMENT,ANALYSISANDMODELINGPart6:Braintool-AnintegratedworkbenchforneuronalmodelingandforrapidprototypingKunoWyler,RolandE.Best,HendrikusW.G.M.Boddeke,DanielStainhauserJuly17,1996AbstractPart5revieweddierentmodelsofneuronsandprovidedasurveyofappropriatesimulationtools.Part6willterminatethisseries‘Fromneurontonetwork:Measurement,analysisandmodeling’withatutorialsectiondescribinganimplementationofaneuronmodelwithMatlab/Simulinkandatypicalapplicationofit.1IntroductionThisnalpartdescribestheimplementationofabiophysicalmodelwithinaprogrammingenvi-ronmentforrapidprototypingofneuronalmodelsusedandcustomizedintheBraintoolproject.BraintoolisaninterdisciplinaryresearchprojectwithintheSwisspriorityprogrammeBiotechnol-ogy(module:Neuroinformatics).Theaimoftheprojectistostudyandsimulatetheinformationprocessinginbiologicalneuralnetworksandtodevelopneuromorphicarticialneuralnetworkalgo-rithmsandarchitecturesofthe3rdgenerationwhicharemorebiologicallyinspiredthantraditionalarticialneuralnetworks.TheresearchisbasedonaclosecollaborationbetweentheInstituteofInformaticsandAppliedMathematicsandtheInstituteofPhysiologyoftheUniversityofBern,Switzerland.Thescienticresearchoftheprojectisfocusedonthreelevelsofmodeling:(1)synapticlevel[1,2,3],(2)celllevel[4],and(3)networklevel[5,6,7,8,9].Tointegratetheresultsofthesedierentsystemlevelsinonemodeloneneedsaexibleprogrammingandsimulationenviron-mentwhichenablesafastimplementationofprototypesofneuronalmodels.Mostoftheavailablesimulationprograms(cf.section4.2ofpart5or[10])donotsupporttheimplementationofnewmodelsorarerestrictedtoaspecicrangeofsystemlevels.Thereforeanotheraimoftheprojectistodevelopanintegratedworkbenchwhichsupportsrapidprototypingofneuronalmodels(rapidneuronalprototyping)andprovidesanextensivesetofanalysisinstrumentsfortheexperimenter.‘Integrated’meansthattheworkbenchsupportssimulationandexperimentationandfacilitatesthecomparisonofphysiologicalexperimentaldatawithsimulationresults(cf.Figure1).ThesimulationenvironmentoftheworkbenchisbasedonMatlab/Simulink[11,12]andtheexperi-mentationenvironmentisbasedonLabView[13].TheuseofgraphicalprogrammingenvironmentslikeSimulinkandLabViewenablesresearcherswithoutbackgroundincomputersciencetoimple-menttheirownmodelsandtousetheminteractivelyina‘plugandplay’manner.Theworkbenchhasanopensystemarchitecture,whichincludesaminimumofxedmodelsandstructures.It1servesasaplatformfortheireasyimplementation.Thus,Braintooldiersfromclassicalsimulationsystemsandcanbeusedmoreliketoadecisionsupportsystemwhichprovidestoolsforacloseinteractionbetweentheoryandexperiment.Toavoidthedevelopmentofisolatedsimulationcodetheworkbenchisbasedontoolboxes,whichcanbesharedbetweenusers.Toolboxesareapplica-tionspeciccollections(libraries)ofreusablecomponents,whichcanbeusedtosetupasinglesimulationorawholeexperimentationenvironment(applications).Everyuserisfreetocustomizehislocalworkbenchbymodifyingcomponentsofatoolboxorbycreatingnewcomponentsorevennewtoolboxes.Ifthesecomponentsareforotherusersofinteresttheywillbeincludedintotheworkbenchprovidedthattheymeetallrequirementsposedina‘Developer’sGuide’(conceptofdistributeddevelopment).Thiswaytheavailablesetoftoolsoftheworkbenchismainlydenedbytheapplicationsandexperimentsofitsusersanditwillgrowwiththeirneeds.Theadvantageofthiscomponentandapplicationorienteddevelopmentisreusabilityofcodeandevolutionarygrowthofthesimulationenvironment.Sothereisnoneedofacompletesystemdesignandaded-icatedteamforprogrammingandmaintenanceofthesystem.Thedisadvantageisperformanceofcodeandorganizationproblemsoftheenvironment.Therstpointcanbehandledeitherbyimplementingtheprototypewithaspecializedsimulatorforneuronalmodels(e.g.Neuron[14])orbyportingthecodetoamorepowerfulcomputationalplatform.Thesecondpointcanbemanagedbydeninganapplicationspecicstructureofworkbenchanaloguetotheoneofarealworkbench.Aworkbenchmakesanumberoftoolboxesavailablecontainingtoolsandcomponentstoundertakeaspecictask.Theexperimentalistusestheworkbenchlikethecraftsman:rst,heanalyseshisproblemandthenhechoosestheappropriatetoolboxandcomponentssuppliedbytheworkbenchandcarriesouthiswork.Ifheneedstoolsorcomponents,whicharenotsuppliedbytheworkbenchhebuildsthemeitherhimselforhegetsthemfromsomewhereelse.Thiscommonimageoftheworkbenchanditspredenedstructurecanbeusedasamentalroad-maptonavigatethroughtheworkbenchtondtheappropriatetoolboxortostoreanewcomponentatitsrightplace.Thispredenedstructurehelpsalsotokeeptheworkbenchinorder.ThenextsectionwilldemonstratethestepwiseimplementationoftheHodgkin-HuxleymodelwithinthesimulationenvironmentoftheBraintoolworkbench.2ImplementationofaHodgkin-HuxleycompartmentThemodelofHodgkin-Huxleydescribestheionicmechanismsunderlyingtheinitiationandprop-agationofactionpotentialsinthesquidgiantaxon.1InthissectionwewillrebuildthismodelwithSimulinkandreproducesomeresultsofthevoltage-clampexperimentsreportedbyHodgkinandHuxleyintheirseminalarticle[15].Simulinkisagraphicalprogrammingenvironmentforthesimulationandtheanalysisofdynamicsy