1. MULTISCALE SYSTEMS MODELLING OFTHE TETRALOGY OF FALLOT 1 1 12 2 23Ron Summers, Tariq Abdulla, Ryan Imms, Jean-Marc Schleich, Guy Carrault, Alfredo Hernandez and Lucile Houyel1Electronic and Electrical Engineering, SEIC, Loughborough University, Leics, UK, LE11 3TUE-mail: R.Summers@lboro.ac.uk Web: http://syseng.lboro.ac.uk2LTSI, University of Rennes 1, Rennes, F-35000, France3Marie-Lannelongue Hospital, Paris, F-92350, FranceIntroductionBetween week 3 and 6 of embryonic development, the human heart morphs from a linear tube to a four chamberedorgan. It is one of the few organs that become functional as it is formed. Remarkably, the conduction system andblood flow both change radically while maintaining cardiac function at every step of development. Heart defects arethe most common type of congenital disorder, severely affecting 6/1000 live births. A number of genes have beenidentified as playing a crucial role in heart morphogenesis. However the mechanisms by which altered genetranscription affects cell signalling, cell behaviour, and tissue-tissue interactions that lead to altered developmentare not well understood. The tetralogy of Fallot is one type of congenital heart disease (CHD), comprising multipledefects, for which a theory of aetiology exists. However this sits within a spectrum of CHD in which one gene actsthrough many mechanisms and can cause one of several diseases. Multiscale modelling, mediated throughinformation models, provides a means to study heart development as a system.Tetralogy of Fallot Complexity of CHDThe tetralogy of Fallot is the most common Several mechanisms are involved in heart development, each of whichcongenital heart defect causing cyanosis, and is are controlled by several genes. CHD commonly involves abnormaldefined as four coinciding anomalies:remodelling of the outflow tract (OFT) which can be caused by a A Pulmonary stenosiscombination of mechanisms, as illustrated below. As the OFT loops B An over-riding aorta, displaced to the rightbehind the atria it septates into the aorta and pulmonary artery and C Ventricular septal defect wedges aligned with the atrioventricular septum. Thus there is a range (always in the membranous septum) of CHDs caused by abnormal degrees of OFT rotation. D Right ventricle hypertrophyWikipediaUser:Wapcaplet ABCD Normal HeartTetralogy of FallotAs the four abnormalities co-occur so frequently, it islikely there is a common cause. One theory is thathypoplasia of the subpulmonary conus leads to bothpulmonary stenosis and a shorter rotation of theOutflow Tract (OFT), which leads to anomalies B , Cand D . Several genes control several mechanisms, which lead to one of several CHDs [1]Multiscale Modelling -9-3The modelling framework encompasses spatial scales from 10 m (protein interactions) to 10 m (the primitive heart tube) and-66temporal scales from 10 s (molecular events) to 10 s (weeks of development). This is illustrated schematically below, left. Theapproach adopted owes much to other methods including those from: systems engineering (e.g. integration technologies andinformation modelling); the world-wide Physiome consortium and the EU-funded Network of Excellence on the VirtualPhysiological Human. Modelling approaches suitable for different levels of scale are illustrated, as well as markup languagespecifications. These enable model interchange, potentially between tools that are suitable for modelling at different scales.-9-6 -4-3 Composite Spatial Scale10 m 10 m10 m 10 mBiosimulation OntologyData SourceAnnotationProtein Cell Tissue Heart Tube Gel ElectrophoresisInteraction BehaviourTransformationMorphogenesisSNAILdecreasedHigh VEGFHigh VEGF//computationconcentrationPRO, GO-MF2+VEGF CABMP2Snail VE CadherinVAR = CalcineurinpVEGFNotchendothelial cellNFATNFATDelta4 Low VEGFVEGF VE-Cadherin2+ CA TGF-beta CalcineurinTGF-beta Wnt /BetaCat Low VEGFp NFAT NFATVEGFSnailpart_of High VEGF Wnt / BMPBetaCat NotchBMP4BMP4 Markup Language SBML CellML CBML FieldMLSNAIL Histochemistry Modelling Pathway Models Stochastic ModelsAgent Based Models Finite Elementdecreased PATO OPB:concentration Approach ODEsReaction Diffusion PDEsReactive AnimationImage Analysis =53 pg ml -1concentration Petri Nets Systems of ODEsCellular Automata 3D ReconstructionGO-CC Boolean Networks Stochastic Petri NetsCellular PottsMultiphysics Simulationendocardial cushionIndependent ContinuantCL PRO, ChEBICL, FMA, GO-CC FMA, EHDA Validation(Proteins, Cells, Structures) derives_into OntologiesGO-MF Cell BehaviourPATO, Mammalian Phenotype Dependent Continuant Segemented MRI(Functions, Roles, Qualities) decreased GO-BPOccurent OPB:area volume(Processes)63volume =3 x 10 mTemporal Scalemembranous part ofFMA, EHDA-6-33 6cardiac septumOMIM / Snomed / AEPC:10 s 10 s10 s10 sMolecular EventsCell SignallingMitosis Heart DevelopmentVentricular Septal Defect Spatial and temporal scales of the multiscale modelling initiative [2] Composite annotation of biomedical data from multiscale sources [2]Reference ontologies applicable to the different levels of scale are illustrated along the bottom of the left-hand figure. These arefurther split between occurents, independent continuants and dependent continuants. Occurents are processes that unfoldthrough time, while continuants are entities that exist in full through a period of time. This provides a clear conceptual divisionbetween the spatial and temporal domains. Annotating models, model components and parameters using well definedontologies enables reuse and integration. But multiscale modelling presents a challenge in that no single ontology can includeterms of the required specificity. A post-coordinated annotation strategy, which allows the combination of terms from multipleontologies, is a partial solution to this issue, and is illustrated above, right. Modelling of morphogenesis provides the furtherchallenge of increased importance of the temporal domain, which is currently less well defined ontologically.Future WorkThere are several important mechanisms in heart development, and each of these can be studied as a multiscale system. Theendocardial cushions are swellings in the early heart tube, which fuse to form the valves and membranous septum, and play a rolein OFT remodelling. Endocardial cushions grow by a process of Epithelial to Mesenchymal Transition (EMT). Cellular behaviourand tissue interaction during EMT can be simulated as Potts models using Compucell3D. Existing models of signal pathwaysinvolved in EMT are modelled as ODEs and are available in Systems Biology Markup Language (SBML). Future plans are to usethe SBML ODE Solver Library (SOSlib) to incorporate reaction networks within Compucell3D and thus determine intracellularconcentrations in a multiscale model. From a chronological perspective, we are using state charts to represent processes and sub-processes in heart development hierarchically. The UML formalism allows the recursive stacking of state machines, and thisapproach neatly matches the problem of modelling in multiple time scales.References[1] F. Bajolle, S. Zaffran, and D. Bonnet, "Genetics and embryological mechanisms of congenital heart diseases.," Archives ofcardiovascular diseases, vol. 102, 2009, pp. 59-63.[2] T. Abdulla, R. Imms, J.M. Schleich, and R. Summers, "Multiscale information modelling for heart morphogenesis," Journal ofPhysics: Conference Series (in press).