- Kevin Burrage (Oxford University and QUT, Brisbane, Australia)
- Gustavo Deco (Universitat Pompeu Fabra)
- Alfons Hoekstra (University of Amsterdam)Multiscale modelling and simulation, lessons learned from COAST (slides)
A few years ago, within the COAST project, we started to develop a multiscale model of a challenging VPH application, namely in-stent restenosis (ISR). This resulted in a multiscale modelling paradigm that we coined Complex Automata (CxA). CxA models can be implemented using our Multiscale Coupling Library and Environment (MUSCLE). I will introduce the main ideas behind CxA, propose a taxonomy of multiscale models, and try to indicate the lessons learned for VPH type of applications.
Multiscale modelling is one thing, but actually executing your multiscale models on computers is quite another ball game. I will show the example of a MUSCLE implementation of a simplified multiscale model for in-stent restenosis. Tightly coupled three dimensional multiscale models will usually require computing power beyond the desktop. This gave rise to the paradigm of Distributed Multiscale Computing, which is currently under development in the MAPPER project.
- Sloot, P.M.A. and A.G. Hoekstra, Multi-scale modelling in computational biomedicine. Brief Bioinform, 2010. 11(1): p. 142-152.
- Hoekstra, A., A. Caiazzo, E. Lorenz, J.-L. Falcone, and B. Chopard, Complex Automata: Multi-scale Modeling with Coupled Cellular Automata, in Simulating Complex Systems by Cellular Automata, A.G. Hoekstra, J. Kroc, and P.M.A. Sloot, Editors. 2010, Springer Berlin / Heidelberg. p. 29-57.
- Falcone, J.-L., B. Chopard, and A. Hoekstra, MML: towards a Multiscale Modeling Language. Procedia Computer Science, 2010. 1(1): p. 819-826.
- Caiazzo, A., D. Evans, J.-L. Falcone, J. Hegewald, E. Lorenz, B. Stahl, D. Wang, J. Bernsdorf, B. Chopard, J. Gunn, R. Hose, M. Krafczyk, P. Lawford, R. Smallwood, D. Walker, and A.G. Hoekstra. A Complex Automata Multiscale Model of In-stent Restenosis. in Lecture Notes in Computer Science. 2009: Springer, Berlin, Heidelberg.
- Hoekstra, A.G., S.F.P. Zwart, M.T. Bubak, and P.M.A. Sloot, Distributed Petascale Computing,, in Petascale Computing: Algorithms and Applications, D. Bader, Editor. 2008, Chapman & Hall. p. 147-164.
- Evans, D.J.W., P.V. Lawford, J. Gunn, D. Walker, D.R. Hose, R.H. Smallwood, B. Chopard, M. Krafczyk, J. Bernsdorf, and A. Hoekstra, The application of multiscale modelling to the process of development and prevention of stenosis in a stented coronary artery. Phil.Trans. R. Soc. A, 2008. 366: p. 3343-3360
- Hegewald, J., M. Krafczyk, J. Tölke, A. Hoekstra, and B. Chopard, An Agent-Based Coupling Platform for Complex Automata, in Computational Science – ICCS 2008, M.T. Bubak, et al., Editors. 2008, Springer: Berlin, Heidelberg. p. 227-233.
- Hoekstra, A.G., E. Lorenz, J.-L. Falcone, and B. Chopard, Towards a Complex Automata Framework for Multi-scale Modeling. International Journal for Multiscale Computational Engineering, 2007. 5(6): p. 491-502.
Some relevant projects:
- Peter Kohl(University of Oxford)What is a model, and why not? (slides)
Indiscriminate use of terminology can be counter-productive for formation of concepts and their communication. Questions such as “What is a model?” or “What is a system?” will raise a multitude of answers, when posed to different people. This lecture aims to propose some common ground and a few basic definitions, from which answers to follow-on questions, such as “What is the best model of … [insert system of your choice]?” or “How soon will we have a complete model of …?” will emerge naturally. It will go on to discuss the multi-scale nature of modelling in bio(-medical) research, illustrate scenarios using the example of research into cardiac electro-mechanical interactions, and end with a praise of failure as the key driver of intellectual insight.
- Bertrand Maury(Université Paris Sud)
- Blanca Rodríguez (Oxford University)
- Paul Watton (Oxford University)The development of computational frameworks to explore the role of the pulsatile haemodynamic environment on the development of cerebral aneurysms for patient-specific arterial geometries (slides)
The formation of a cerebral aneurysm is a complex mechanobiological process which is not yet clearly understood. However, the mechanisms that give rise to its development involve the complex interplay between the local mechanical stimuli acting on the arterial wall and the biological processes occurring at the cellular level. The inner surface of the artery is comprised of endothelial cells (ECs). Wall shear stress (WSS) is sensed by the EC and transduced into biochemical signals which activate signalling pathways to control its functionality and the functionality of the artery. The structure of the artery is continually maintained by fibroblasts and vascular smooth muscle cells. These cells secrete connective tissue and matrix degrading enzymes and may up/down-regulate expression levels in response to deviations of mechanical stimuli, e.g. their state of loading from normotensive levels.
Recently, Watton et al (2009a, 2010) proposed a framework to couple the evolution of a saccular cerebral aneurysm to the haemodynamic environment. Briefly: the model utilises a realistic constitutive model of the arterial wall and links elastin degradation to deviations of WSS and spatial WSS gradients from normotensive levels; collagen remodels to maintain an equilibrium level of strain and the mass of collagen adapts to simulate fibroblasts responding to deviations of stretch from normotensive levels. However, the model considered idealised arterial geometries and linked arterial growth and remodelling (G&R) to steady flow parameters. In this study we extend the framework to patient-specific arterial geometries and link arterial G&R to mechanical stimuli that arise due to the pulsatile nature of the blood flow.
A patient-specific cerebral aneurysm case is identified from clinical imaging data. The imaging data of the cerebral vasculature is automatically segmented and the geometry is subsequently manipulated with ANSYS ICEM: the aneurysm (located on the right internal carotid sinus artery) is removed and replaced by a short cylindrical section which is reconnected to the upstream and downstream arterial sections so that the surface gradients are continuous. It is within this inserted section, hereon referred to as “aneurysmal section”, that the formation of a new aneurysm is simulated. A rigid-wall approach is adopted to solve the pulsatile flow using physiological boundary conditions for the right internal carotid, and middle/anterior cerebral arteries. This enables G&R of the aneurysmal section to be explicitly linked to physiologically realistic haemodynamic stimuli. In addition, a quasi-static approach is used to obtain the geometry of the aneurysmal section at systolic and diastolic pressures, enabling G&R to be explicitly linked to the magnitude of the cyclic deformation of vascular cells. This is the first patient-specific model of cerebral aneurysm evolution that incorporates a realistic constitutive model of the arterial wall and explicitly links G&R to the pulsatile mechanical environment.
The framework provides the basis for further investigating and elucidating the aetiology of the disease. Further sophistications are required. For example, one natural development would be to model the signaling pathways linking the functionality of the cells to the mechanical and chemical environment. Such models could be integrated into conceptual models of aneurysm development (Watton et al 2009b) prior to integration into more sophisticated computational frameworks (Watton et al (2010)).
Elements of Multiscale nonstandard simulation techniques (slides)
Standard mathematical models of complex dynamical biological processes include elements of ordinary differential equations to capture, for example, biochemical kinetics coupled with continuum partial differential equations models that represent spatial elements such as transport and motion. However, there is increasing interest in nonstandard simulation techniques that capture, for example, discrete stochastic subcellular behaviour or spatial heterogeneity in media via the concept of fractional derivatives. My challenge in 50 minutes is to give an overview of some of the important multiscale modeling and simulation issues associated with these nonstandard approaches.
How Local Network Oscillations lead to Functional Networks during Rest
Spatiotemporally organized spontaneous low-frequency (< 0.1 Hz) fluctuations have been revealed by the blood-oxygenation level-dependent (BOLD) fMRI signal during rest. Indeed, in the absence of a task, significant correlations between distinct anatomical regions are found. These correlations, referred to as functional connectivity (FC), yield large-scale maps constituting so-called resting-state networks (RSNs). Furthermore, direct measurements of the neuronal activity have revealed similar large-scale correlations, particularly in the slow fluctuations of the power of local field potential gamma frequency range oscillations. Nevertheless, the origin of this highly structured slow dynamics and its relationship with neural activity, particularly in the gamma frequency range, remains largely unknown. To address these questions, we defined a model of brain neural activity taking into account the long range connectivity together with their corresponding conduction delays and instantiating sustained gamma oscillations in the dynamics of its local nodes. We apply the model to the macaque and human measured structural connectivity. In the human case, we search for parameters such that the model best reproduces the human empirical FC obtained at the same nodes. The best agreement is found in a region of the parameter space where the network is globally in an incoherent state but where partial clusters of nodes tend to synchronize. Inside such clusters, the BOLD signal between nodes is found to be correlated, instantiating then RSNs. Between clusters, patterns of positive and negative correlations are found, as in experiments. These results are found to be robust to a number of model parameters.
Multimodel description of the human lungs (slides)
The respiratory system realizes the transfer of oxygen from the outside air to the alveolar membrane, through which it diffuses onto the blood. As pure diffusion is far from being sufficient to realize that transfer, most of it is of advective type, and this advection is triggered by inflation-deflation cycles of the paremchyma.
The mechanical part of the lungs can then be seen as a tree-like domain (conducting airways) embedded in an elastic medium. The flow in the upper part is inertial (incompressible Navier-Stokes equations), whereas inertia can be neglected for deeper branches (Stokes equations), which allows to use Poiseuille’s law for each branch, and consequently Darcy like equations on the corresponding subtrees.
We aim at addressing the delicate issues in terms of theory, numerics, and modelling, raised by the coupling of those models (Navier-Stokes, Darcy equations on a network, elasticity equations).
Multiscale modelling and simulation of drug-induced effects on the heart (slides)
In 1962, Denis Noble published the first mathematical model of a cardiac cell action potential based on the Hodgkin-Huxley formulation.
Since then, computational cardiac electrophysiology has developed into a mature discipline in which advanced computational, mathematical and engineering techniques are used to investigate heart rhythm mechanisms in health and disease.
The causes of cardiac arrhythmias are numerous and include disease, mutations and also drug-induced abnormalities in ionic properties. Of
particular concern for regulatory agencies, pharmaceutical industry and society is the fact that certain drugs, in particular those not
designed to affect the heart, can exhibit cardio-toxicity (i.e. unwanted side effects), which can put patients with otherwise healthy
hearts at risk of developing lethal arrhythmias.
In this presentation, we will describe the state-of-the-art in multiscale modelling and simulation of ventricular electrophysiology,
and we will illustrate their use in the investigation of drug-induced abnormalities in heart rhythm mechanisms. The ultimate goal of the
research described here is to contribute to the improvement of the diagnosis and treatment of cardiac arrhythmias to reduce the burden
they impose to society.
- Watton PN, Selimovic A, Raberger NB, Huang P, Holzapfel GA, Ventikos Y, (2010) Modelling Evolution and the Evolving Mechanical Environment of Saccular Cerebral Aneurysms, Biomechanics and Modelling in Mechanobiology (in press), DOI: 10.1007/s10237-010-0221-y
- Watton PN, Raberger NB, Holzpfel GA, Ventikos Y (2009a) Coupling the Haemodynamic Environment to the Evolution of Cerebral Aneurysms: Computational Framework and Numerical Examples, ASME J Biomech Eng, 131(10).
- Watton PN, Ventikos Y, Holzapfel GA, (2009b) Modelling the growth and stabilisation of cerebral aneurysms, Mathematical Medicine and Biology, 26(2):133-164
- Jonathan Cooper (Oxford University)
- James R. Dalton and Albert Mascarell (Universitat Pompeu Fabra)
- Yves Martelli (Universitat Pompeu Fabra) and Keith McCormack (University of Sheffield)
- Stefan Zasada (University College London)
Languages for Multiscale Modelling and Simulation (slides)
This tutorial will cover markup languages designed to express multiscale models, both their uses and limitations. It will start by describing the general concepts behind such languages, why they have been (and are being) developed, what problems they are designed to solve, and how they solve them. This will include an overview of the most common existing languages, comparing their features and suitability for different tasks.
Next, we will examine one such language, CellML (http://www.cellml.org) in more depth, looking in detail at the constructs available in the language and hence what it can be used to model. This section will cover both best practices for developing models in CellML, and available tools for working with CellML.
Finally, the tutorial will conclude with a live demonstration of how CellML can be used for a particular multiscale modelling problem: drug action on the heart. Two tools will be demonstrated: OpenCell (http://www.cellml.org/tools/opencell) and Chaste (http://www.comlab.ox.ac.uk/chaste).
Activ8 is a new online multi-functional tool that provides a “3-D world” where researchers can shape their own working environment, integrate data, access communication tools, develop projects, schedule events, access databases, build workflows, run applications, track new developments and network with fellow scientists. Furthermore, Activ8 provides a means of modelling and visualising complex biological information via a top down approach, through a cascade of knowledge 3-D environments, from the organ to the tissue, from the cell to sub-cellular compartments, from the protein to the gene, and from pathways to interactions. This multi-scale approach is based on XML standards, the integration of geographically dispersed data sources, and the inclusion of specific protocols established in each layer. Our aim is to provide integrative biomedical models that can play a role in personal health care, benefiting both the clinician and patient alike, with the current focus of Activ8 on the modelling of cystic fibrosis – although soon to be extended to neurological disorders such as cerebral palsy.
This tutorial will provide a live demonstration of Activ8 (http://www.o2hlink.com), starting with some general functionalities and integrated modelling tools, and proceeding onto the multi-scale visualisation of the different elements involved in cystic fibrosis, and the way in which the disease can be understood in each biological layer.
The VPH NoE ToolKit: An Update on Progress, October 2010 (slides)
The Virtual Physiological Human (VPH) ToolKit is a repository for computing Tools, Models and Data designed to assist the VPH community in its work to build physiological computational workflows for medical research and clinical decision support. This talk explains the progress being made by the VPH NoE in establishing the ToolKit as a sustainable resource that will endure into the foreseeable future.
Accessing European Computing Platforms (slides)
This tutorial will look at the computational resources available in Europe to support VPH style research. We will examine the different resources available, and look at which are suitable for which types of problem, and then go on to look at the lightweight mechanisms put in place by the VPH NoE to allow VPH researchers to apply for access.
Next, we will cover the tools and technologies used to access these computational resources, including grid middleware such as Unicore and Globus, dealing with digital certificates, and higher-level tools such as the Application Hosting Environment. This section will then briefly look at how to compile codes on HPC resources, and some of the options for parallelizing an algorithm.
Finally, we will look at the workflow tools available to VPH researchers that allow for multiscale simulations on EU resources, and where researchers can get support in order to use the available resources. We will conclude by looking at some of the forthcoming infrastructure developments that will impact VPH researches.