Session 1 Speakers: IMAG Integrative Modeling: Challenges in Modularity

Biological systems are inherently modular but the modules are highly connected, with multiple feedbacks. Automated reconstruction from archived modules requires adherence to standards, the use of ontologies, and computer-recognition of relationships between processes. Even more challenging is the automation of the processes of reducing model complexity for using modules in higher level, more speedily computable versions. Success in simplification means a reduction in robustness, defining robustness as the generic ability to respond appropriately to changes, thus recognizing that only certain types of responses can be preserved while reducing model complexity. The residual fundamental problem then is, when inadequacy is identified, how to unreduce the model efficiently to reincorporate only the critical missing characteristics.

 

Name Email Talk Title
Jim Bassingthwaighte jbb2@u.washington.edu Modular Construction of Multiscale Models
Dan Cook/John Gennari dcook@u.washington.edu Exploiting the biophysical semantics of biosimulation models
Poul Nielsen/Peter Hunter p.hunter@auckland.ac.nz Markup languages, model repositories & toolkits for the EuroPhysiome/VPH project
Marco Viceconti viceconti@tecno.ior.it VPHOP Hypermodelling technology: a peer-to-peer conceptual architecture for the modular composition of multiscale predictive models

Back to Minisymposium_planning.

ABSTRACTS

  • Bassingthwaighte:Automating Modular Construction of Multiscale Models
James B Bassingthwaighte, Howard J Chizeck, Erik Butterworth, Gary M Raymond
Depts of Bioengineering and Electrical Engineering, University of Washington, Seattle WA 98195-5061
Success in automating automating the construction of integrated models from databases of preconstructured modules (= modular model forms) would allow the extraction of thoroughly validated modules from well-curated databases, with assurance of clear documentation by acknowledged experts. Competing modules should similarly be available. To pull selected modules into a higher level model requires common ontologies, complete equations with boundary and initial conditions. completely defined units, and running examples. This requires serious collaborative planning, at the national and international levels. Total success in very general way is unlikely. The problems start with the fact that models stored in databases such as CellML, SBML or ours at Physiome.org or others, have not adhered to a common ontology. Having multiple synonyms for variable and parameter names would help, at some risk of loss of uniqueness. Suggestions for implementing such capabilities is the topic of this discussion.
  • CookExploiting the biophysical semantics of biosimulation models — tools for automated merging and encoding of multiscale/multidomain models Daniel L. Cook, John H. Gennari, Maxwell L. Neal
Currently, the merging of validated physics-based modules into multiscale (molecules to organisms) and multidomain (chemical kinetics, fluid dynamics, etc.) biosimulation models depends on manual, code-level integration — methods that scale poorly to the scope of the physiome. Providing computational assistance, or outright automation of model merging must rely on: (1) computable knowledge of the biophysical properties and property dependencies that models encode as, respectively, variables and equations, and (2) a corresponding bioinformatics architecture and tool set that can merge and encode models across disparate modeling languages while resolving pervasive curatorial errors. Toward these goals, we introduce (1) the Ontology of Physics for Biology (OPB) — an ontology of biophysical and systems dynamical knowledge, (2) Semantic Simulation (SemSim) models — light-weight ontologies that encode the biophysical semantic content of individual models in terms of the OPB and other knowledge resources, and (3) BioSem Builder — software that semantically merges SemSim models and re-encodes the merged models into a variety of biosimulation languages. We anticipate that these bioinformatics knowledge resources and tools will greatly accelerate and facilitate automatic, or semi-automatic composition of patient- and pathology-specific biosimulation models.

Media:SIAM08-MSM-Cook.ppt

 

  • Nielsen/Hunter:Markup languages, model repositories & toolkits for the EuroPhysiome/VPH project
P Nielsen(1), R Britten(1), A Garny(2), J Lawson(1), C Lloyd(1), J Marsh(1), A Miller(1), D Nickerson(3), P Noble(2), T Yu(1), P Hunter(1,2)
(1)Auckland Bioengineering Institute, University of Auckland, (2)Department of Physiology, Anatomy & Genetics, Oxford University, (3)Division of Bioengineering, National University of Singapore.
To facilitate model reuse among European researchers in a new Network of Excellence (NoE) for the EuroPhysiome or "Virtual Physiological Human" (VPH) project, two XML markup languages for encoding biological models, CellML (www.cellml.org) & FieldML (www.fieldml.org), are being further developed. CellML deals with models of so-called ‘lumped parameter’ systems, where spatial effects are averaged, and typically involves systems of ordinary differential equations and algebraic equations. FieldML addresses the spatial variations in cell or tissue properties where the models typically rely on partial differential equations. The two standards can be used together. These languages, which define the structure of a model, the mathematical equations and the associated metadata, enable (i) automated checking to ensure consistency of physical units used in the model equations, (ii) models developed by different groups to be combined using commonly agreed ontological terms within the metadata, (iii) models to be modularized and used in libraries to make it easier to create complex models by importing simpler ones. Model repositories based on these standards and implementing a wide variety of models from peer-reviewed publications have been developed (www.cellml.org/models) and open source software tools for creating, visualizing and executing these models are currently available (www.cellml.org/tools) and under continuous development. This talk will describe the new VPH NoE, new developments in CellML and also briefly describe recent progress on FieldML.

Media:NielsenHunter_Montreal.ppt

 

  • Marco Viceconti:VPHOP Hypermodelling technology: a peer-to-peer conceptual architecture for the modular composition of multiscale predictive models
Marco Viceconti1, Fulvia Taddei1, Debora Testi2, 1Istituti Ortopedici Rizzoli, Bologna, Italy, 2 CINECA supercomputing centre, Bologna, Italy
The Osteoporotic Virtual Physiological Human (VPHOP) is a large-scale international project that will run in the period 2008-2012. The scope is to develop a patient-specific multiscale modelling technology capable of predicting with high accuracy the absolute risk of bone fracture in osteoporotic patients. The presentation will describe the conceptual architecture of this model of models, a.k.a. hypermodel, and some of the implementation strategies we plan to adopt to solve this complex problem. In particular we shall describe the idea of an agent-based architecture that leverages of digital library services for data management to ensure neutrality and easy interoperability, and an adaptive communication protocol to solve the problem of hypermodels combining loosely and tightly coupled models.

Media:Pres_viceconti_SIAM2008.ppt‎

Table sorting checkbox
Off