Goals and Objectives:
The overall goal of WG4 is to discuss cell level modeling in the model sharing context. Most of our activities over past two years have been focused on identifying key issues that are crucial for implementation of shareable models. While various groups around the world have put a lot of efforts into model sharing, there remains a lot of work to do in order to define model sharing standards for biomedical community. The purpose of the discussions within this group is to research what is necessary to establish such standards and whether it is possible/practical at all to standardize models.
White Paper Discussion
Frameworks for Multi-cell Aware Multiscale Integration Jim Sluka, James glazier, Maciej Swat
Conjecture: future, scientifically useful, multiscale models will include new classes that rely on relational grounding Anthony Hunt Tony_Hunt_Draft_MSM_Position_paper_v3.doc
Monday July 27, 2009 4-5pm EDT - Benjamin Zaitlen, Indiana University
TITLE: Cell Behavior Ontology(CBO) - relevance to biological modeling, challenges, opportunities and CBO workshop summary
ABSTRACT: Ontologies are logical structures which provide formal descriptions of (scientific) concepts, hierarchies of terms with agreed meanings and relationships. Some examples of ontology-related projects for biology include the Gene Ontology, MGED, SBO and KiSAO. When successful (as with the Gene Ontology which is a major component of the Genomic revolution), they enable seamless integration of software and unambiguous specification of meaning. (read more on Ontologies)
The Cell Behavior Ontology (CBO) is intended to provide such seamless integration as well as sharing between applications that seek to model biology at the cell and tissue level. In this presentation we will overview how CBO can impact biomedical cell- and tissue-level modeling, why ontologies are enabling technologies and we will summarize first Cell Behavior Ontology Workshop that was held in Bethesda, MD, on May 4-6 2009. The goal of the workshop was to complete a top-level draft ontology describing cell behaviors which is acceptable to the broad bioscience community. After CBO version 1.0 we hope to begin development of implementation independent Cell Behavior Markup Language (CBML) which will isolate model description from model implementation. If successful the CBML will greatly enhance model sharing, model integration and model curation and collaboration between researchers.
Presentation: Media: CBO_IMAG-MSM.ppt
CBO wiki: http://cbo.compucell3d.org
Monday December 1, 2008, 4pm EST - Maciek Swat, Biocomplexity Institite, Indiana University, Bloomington ,IN, USA
From regulatory networks to tissue level phenomena – multi-scale modeling of biological systems using CompuCell3D and SBW
Mathematical modeling and computer simulation have become crucial to biological fields from genomics to ecology. However, multi-cell, tissue-level simulations of development and disease have lagged behind other areas because they are mathematically more complex and lack easy-to-use software tools that allow building and running in-silico experiments without requiring in-depth knowledge of programming.
Biologically relevant simulations should capture key cell-level behaviors, providing a phenomenological description of cell interactions without requiring prohibitively detailed molecular-level simulations of the internal state of each cell. While an understanding of cell biology, biochemistry, genetics, etc. is essential for building useful, predictive simulations, the hardest part of simulation building is identifying and quantitatively describing appropriate subsets of this knowledge. In the excitement of discovery, scientists often forget that modeling and simulation, by definition, require simplification of reality.
In this talk I will give an overview of the approach taken by IU and UW groups to deploy truly multiscale simulation environment for building, testing and running multi-cell models which require user-selectable modeling resolution, from sub-cellular compartmental models to continuum models of tissues.
I will show examples of concrete modeling situations where the ability to model at the sub-cellular level is a prerequisite for building realistic tissue-level models.
In the context of multi-scale multi-cell modeling I will briefly discuss our needs to develop standard way of describing "what a cell is", "what it does" and "what phenomenological properties it has" in a way that would be independent from underlying modeling methodology. Link to the presentation can be found here: []
Cell Behavior Ontologies
The purpose of this discussion is to create an ontology that would address needs of modelers who treat cells phenomenologically. In contrast to existing cell ontologies most of which concentrate on e.g. different cell types, cell classification from physiological point of view etc. , cell behavior ontology would attempt to categorize cell behaviors using "common sense" approach. For example we know that cells move, stick to each other, grow, divide, die, secrete signals absorb signals etc. All such behaviors should be properly put in the cell behavior ontology.
Although we have ideas as far as what new cell behavior ontology should contain, we need to rigorously define what aspects of cell biology we will deal with and which aspects we will have to ignore (at least at the beginning) We also need to decide what we expect to happen when cell ontology will be created. Is it going to be used as a base for model sharing efforts? Are we going to develop new ML based on it? etc ...
Ontologies are logical structures which provide a formal description of (scientific) concepts. An ontology is simply a hierarchy of terms with agreed meanings and sets of subterms and modifiers which can be applied to each term. Ontologies are unglamorous and tedious to develop, but also crucial for the development of software applications in their subject areas. In some cases the development of an ontology has revolutionized a scientific discipline. E.g. ontologies like the Gene Ontology (GO) allow the hundreds of programs which process genetic-sequence information to communicate with each other seamlessly—the output of any program serving as the input for another. Without GO each program would have to handle the separate formats produced by each other program and interpreting the huge masses of sequence data produced by contemporary genomics would be simply unfeasible. Similar standards are crucial to Geographical Information Systems for map-making and weather forecasting, Molecular-Dynamics simulations in biochemistry,… Where such standards are lacking, as in the representation of medical records, chaos reigns.
While numerous ontologies exist for the description of both continuum biological tissues and subcellular phenomena (like GO or the Cell Phenotype Ontology), no ontology describes such simple cell behaviors as cell division, cell movement, cell death or changes of cell shape. The lack of agreed-upon terms means that anyone wishing to model a process involving these behaviors (e.g. cancer growth, regeneration, wound healing, or normal biological development) must define these terms from scratch. Someone reading a paper and wishing to reproduce or adapt its results needs first to figure out what the model actually was. Since most scientists make their ontological decisions tacitly, reconstructing their ontologies is difficult. If, as often happens, their ad hoc ontologies were not logically consistent, reconstruction can be impossible.
The only thing approximating an ontology of cell behaviors available at the moment is an eXtended Markup Language (XML) called CC3DML for specifying multi-cell models for the modeling environment (CompuCell3D, CC3D). Because of the ad hoc process of CC3DML’s development, multi-cell models written in CC3DML cannot run in other modeling environments and do not optimally integrate with subcellular descriptor languages like CellML and SBML. This idiosyncrasy was not a problem five years ago, but is becoming more serious as CC3D finds more users and other scientists develop alternative multi-cell developmental-biology modeling environments. Ideally a model written in one should run in any other, allowing the user to separate implementation-specific artifacts from the fundamental behavior of the model. The first step towards developing such an XML is to develop a comprehensive ontology for the class of problems.
Discussions in the IMAG project, one of whose focus areas is the development of ontologies, show that we have the crucial community buy-in and agreement to finally develop a functional and widely-used cell-phenomenology ontology. The time has come to develop a draft ontology. This work would include literature review, writing, coordination with other modelers and experimentalists interested in cell behavior and development and the organization of several small intensive workshops to brainstorm and refine the ontology. The deliverable will be a draft ontology in the OWL format (the most accepted current software for ontology development) and a specification document for the ontology in text.
An initial workshop (lasting 2-3 days) would focus on agreeing on the top and second tier definitions of cell behaviors and reviewing any plans that would overlap with the proposed ontology.
Links to existing resources
Below one we include links to existing cell ontologies. As mentioned above they do not directly address needs of this working group. They may contain useful bits and pieces that we may reuse in our cell behavior ontology
Existing cell-related ontologies (to be updated):
Relevant Publications: CELL BEHAVIOR DATA BASE (2): PREPARATION FOR BUILDING A CELL BEHAVIOR "ONTOLOGY" DAN MARINA Zoological Science vol: 22 issue: 12 page: 1450 year: 2005
Cell Behavior Ontology Workshop May 2009, Bethesda,MD
After phone conversation with Herbert Sauro we have decided to TRY to organize the Cell Behavior Ontology (CBO) Workshop during the week of May 4th or the week of May 11th. The list of possible keynote speakers is shown below:
James Binkley (U.Washington)
John Gennari (U.Washington)
Dan Cook (U.Washington)
Herbert Sauro (U.Washington)
Michael Hucka (Caltech)
Mark Musen (Stanford)
Jim Bassingthwaighte (U.Washington)
Frank Bergmann (U.Washington)
Ryan Roper (U.Washington)
Lucian Smith (U.Washington)
Nicolas Le Novère (EMBL-EBI)
James A. Glazier, Benji Zaitlen, Randy Heiland, Maciej Swat (IUB)
Peter Hunter and/or designee (U. Auckland)
Tim Newman (ASU)
Andreas Deutsch (Dresden)
Paulien Hogeweg and/or designee (Utrecht)
Rod Smallwood (Nottingham)
Wayne Brodland (Waterloo, Canada)
Corneljis Weijer (Dundee, Scotland)
Roeland Merks (University of Ghent)
Imran Shah (EPA)
Georg Luebeck (FHCRC Seattle)
Frank Hartel (CaBIG)
George Komatsoulis (CaBIG)
Vittorio Cristini (U.Houston)
Lance Munn (Harvard Medical School)
NIH/NSF representation (Peter Lyster, Dan Gallahan, Grace Peng)
In general we are trying to invite scientists who are experts in developing ontologies per se and scientists who have experience in cell-level biology - both experiment and modeling. If we can put have interaction between those 3 categories of scientists we will be able to meet the deliverables.
Georg Luebeck: "It makes sense to link ontologies to applications in as far as applications usually address relevant biomedical problems and therefore provide a framework for testing and validation. However, there is also the danger that such ontologies are too much driven by physical abstractions that do not jive with molecular cell biology. I very much would like to see an involvement of experimental colleagues who study cell behavior under a variety of physiological conditions (Celeste Nelson and Donald Ingber come to mind). Perhaps, we should also invite a few experts in cell cycle regulation, differentiation, cell transformation, wound healing(Bill Carter FHCRC) etc. to this or a future cell ontology meeting."
Herbert Sauro:"The thought is to have a two day workshop involving talks and discussion with maybe >20 or more people followed by a one day more intense session to actually put together a first draft of the ontology. The third day may involve fewer people. The workshop should definitely aim to get something concrete done, that is a first draft of an ontology."
Purpose and deliverables
1. To discuss what constitutes Cell Behavior Ontology (CBO). A.k.a. what aspects of cell biology should/should not be addressed by CBO
2. Review of existing ontologies related to cell biology. Present one example (does not have to be biological) where ontologies proved to be essential for the completion of the overall project. This should be a crystal clear talk that would avoid jargon and explain in laymen-terms reasons why ontologies are useful.
3. There was a suggestion that part of the workshop should be dedicated to the discussion on what we are trying to accomplish with CBO. One goal that probably most of us have in mind is implementation independent language for cell level modeling. Is this a reasonable goal for the CBO effort (not just the workshop)?
4. Deliverable : White paper summarizing the workshop. The paper should contain at the very least a draft of CBO. Preferably the paper would contain CBO v1. It is clear that CBO will evolve but to start this evolution we NEED to have a version of CBO.
Software for ontology building
Herbert Sauro: "Talking to people at UW and our own experience from involvement with the systems biology community, I would recommend an online community based ontology manager (maybe the online protege software), something similar to what is used for the SBO ontology in the UK. However with caBIG being in attendance maybe the bioportal project would offer another option. Part of the success of SBML/CellML and associated ontologies and standards stems from the fact that they were grass-roots efforts, bottom-up rather than top-down."
BioModels Database is the largest repository of computational cell models today. As of July 2009, there are over 410 models in the database, of which 216 are fully human-curated, and this number is growing continually. BioModels Database permits models to be uploaded in SBML and CellML format and downloaded in many different formats, including SBML, CellML, SciLab, XPP and BioPAX.
Anyone who has tried to extract a model directly from a published paper will know how difficult it is. The curators for BioModels Database report that well over 95% of published models would not run as-is; the underlying causes are often simple typographical errors given in the model equations or missing parameter values, but sometimes the problems are deeper. The curators often have to contact the authors of the published models in order to reconcile differences in the models described in their papers and the claimed simulation results claimed. The experiences of BioModels Database highlights the need for journals to accept computer readable formats of models when authors submit papers describing models.
An additional, unique feature of BioModels Database is that all models are annotated by hand and linked to relevant data resources, such as publications, databases of compounds (e.g., ChEBI, UniProt) and pathways (e.g., KEGG), controlled vocabularies (e.g., GO), etc. This makes every model in BioModels Database richly cross-linked and semantically annotated.
The CellML site has a very large collection (258 models) of models encoded in CellML. The repository is currently in the process of being curated. By December 2008, about 180 have been curated by the Auckland team organized by Catherine Lloyd.
This site at the University of Washington contains about 250 models, some from the literature and others from the experimental programs of Jim Bassingthwaighte and Carl Goresky, and msny others as modules or teaching models. The models are all curated and can be run from the website or downloaded. They are written mainly in MML (Mathematical modeling language) and run under JSim (for linux, MAC, Windows). The site is under renovation in winter 2009. For teaching purposes thee are tutorial sets of models designed to assist learning by starting with simple models and going to the more complex ones.
CompuCell3D is an open source modeling environment and pde solver, primarly used to study cellular behavior. CompuCell3D was initially developed to help researchers (biologists and physicists) model materials and tissues without having to dedicate resources to developing code that will reproduce existing software. The code is built on C++ with a Python wrapper. Users can easily define model parameters in an XML file and run the simulation with our GUI, the CompuCell Player. The GUI gives users an easy interface to complex models as well as real time visualization of their simulations. CompuCell3D is freely downloadable from: http://www.compucell3d.org/
TRND is an integrated workflow wherein user-supplied gene expression data generates transcriptional regulatory networks and derives their biological implications. TRND uses a built-in database of experimentally validated gene-transcription factor regulatory interactions and a non-linear dynamical systems analysis package for discovering transitions in cell behavior supported by a transcription-translation-post-translation process network.
Karyote is an integrated set of modules for building and simulating ordinary differential equation models of single cells and arrays of interacting cells. Karyote is hierarchical so that models can accommodate intracellular compartments for arrays of interacting cells. It allows automatic integration of simpler models to create more complete ones and has a set of example models.
JSim: A platform for the modeling analysis of data
JSim is a modeling system for designing and coding models and for applying them to the analysis of data, facilitating hypothesis testing and the parameterization of data. JSim is Java based; the MML (mathematical modeling language) code for the models is parsed and compiled into Java for execution. Fortran, C, and matlab code can be run under it. The system provides for multiple graphs (2D or 3D) for outputs, comparisons af two or more models in the same project file, uses of multiple parameters sets for varied model configurations, 8 different ODE solvers and 3 PDE solvers, 6 optimization routines for automated data fitting by models, confidence limit calculation, and sensitivity analysis. The project file can incorporate many experimental data sets for analysis. (submitted by J Bassingthwaighte 22jan09)
SBW: Systems Biology Workbench
The Systems Biology Workbench (SBW) is a software framework that allows heterogeneous application components--written in diverse programming languages and running on different platforms--to communicate and use each others' capabilities via a fast binary encoded-message system. Numerous plugins exist for the SBW, including simulators, model editors, visualizers and analysis systems. All plugins accept (or export) standard SBML.
TinkerCell: Engineering platform for building and testing cellular circuits
TinkerCell is an extensible platform for editing and simulating cellular networks. Users can operate the software at different levels including graphical point and click or via an interactive console. It has the ability to interact with databases such as RegulonDB to provide the user with parts for the construction of new or existing networks from which the DNA sequence can be generated for later fabrication. TinkerCell is cross platform and written in C++. A Python console is provided for interactive control.
CompuCell3D Training Workshop Aug 13-Aug 17 2007, Bloomington, Indiana
Biocomplexity Institute (Indiana University) is organizing first CompuCell3D training workshop. Application deadline is Jul 15th 2007.
For more information please see workshop poster
- or visit