Actions

Rule-Based Modeling of Biochemical Systems: Difference between revisions

From Santa Fe Institute Events Wiki
Line 56: Line 56:


* [http://genetics.uchc.edu/faculty/blinov.htm Michael L. Blinov (U Connecticut Health Ctr)]: '''SBML Features Necessary for Rule-Based Modeling of Reaction Networks'''
* [http://genetics.uchc.edu/faculty/blinov.htm Michael L. Blinov (U Connecticut Health Ctr)]: '''SBML Features Necessary for Rule-Based Modeling of Reaction Networks'''
:: ''Abstract'': We will discuss SBML features that are necessary for rule-based modeling. We will discuss how to encode in SBML: 1) Multi-state multi-component species expressed as typed graphs with explicit representation of the connectivity of species’ components, 2) Reactions in the form of graph transformations, 3) Patterns that select groups of species having user-specified properties, 4) Rules that define transformations of sets of species selected by some pattern, 5) Species represented as hierarchical graphs of an arbitrary degree of nesting.
:: ''Abstract'': I will discuss SBML features that I believe are necessary for rule-based modeling. I will present a proposal for how to encode in SBML: 1) Multi-state multi-component species expressed as typed graphs with explicit representation of the connectivity of species’ components, 2) Reactions in the form of graph transformations, 3) Patterns that select groups of species having user-specified properties, 4) Rules that define transformations of sets of species selected by some pattern, 5) Species represented as hierarchical graphs of an arbitrary degree of nesting.


* [http://sysbio.med.harvard.edu/faculty/sorger/ Will Chen (Harvard)]: '''Phospho-dynamic Experiments and Computational Modeling of the ErbB Pathways'''
* [http://sysbio.med.harvard.edu/faculty/sorger/ Will Chen (Harvard)]: '''Phospho-dynamic Experiments and Computational Modeling of the ErbB Pathways'''

Revision as of 19:44, 11 June 2007

The use of rules to represent and simulate the interactions of molecules in biochemical regulatory networks [1] is an emerging area of systems biology that promises to change the way these networks are modeled and understood. The purpose of this proposed workshop is to bring together representatives from major research groups active in this area for the purpose of exchanging ideas, discussing applications of rule-based modeling methods, surveying existing capabilities, mapping out where the field is headed, and developing standards that will promote the exchange of models and the development of new software tools. As part of this, one pratical goal will be to discuss and (hopefully) settle on a draft proposal for representing multicomponent species and complexes within the framework of SBML.

Workshop Organization

The workshop organizers are:

This workshop is made possible thanks to generous support from the following organizations:

Preliminary Workshop Schedule and Venue Information

We will begin with a reception on Wednesday June 13 (to be held at the Hotel Santa Fe), followed by two full days of meetings at the Santa Fe Institute on June 14 and 15.

All talks will be 25 minutes followed by 5 minutes for questions.

Here's a preliminary agenda.

Background Readings

If you have time for nothing else, we strongly recommend reading the following before coming to the meeting:

  • Hlavacek WS, Faeder JR, Blinov ML, Posner RG, Hucka M, Fontana W, Rules for modeling signal-transduction systems, Science STKE, 18 July 2006 PDF. This is a a comprehensive review of rule-based modeling written by the organizers.
  • Kohn, KW, Aladjem, MI, Kim, S, Weinstein, JN, Pommier, Y "Depicting combinatorial complexity with the molecular interaction map notation" Mol Syst Biol 2006 PubMed Abstract

The following are important for the discussions about extending SBML to support this class of modeling:

  • Michael Blinov et al.'s 2004 proposal for SBML Level 3 extensions for multicomponent species PDF
  • Andrew Finney's 2004 proposal for SBML Level 3 extensions for multicomponent species PDF
  • The SBML Level 2 Version 3 public prerelease specification PDF

Additional readings for context and background (speakers - please suggest a paper, ideally current and technically detailed):

  • Blinov, Yang, Faeder & Hlavacek, "Graph theory for rule-based modeling of biochemical networks," Lect. Notes Comput. Sci. 4230, 89-106 [2]
  • Lok & Brent, "Automatic generation of cellular reaction networks with Moleculizer 1.0", Nat. Biotech. 23(1):131-136, 2005. PDF


Attendees and Abstracts

The following are the confirmed attendees and the current abstracts of their presentations.

Abstract: This talk will detail my experiences writing a converter that converts between BNG models and Moleculizer models. Both programs are examples of chemical reaction network simulators that take input in the form of template based rules describing the conditions and rates under which new complexes form. In spite of their similarities, there are several important differences in their structure, which represent different schools of thought as to how rules should be specified. This talk will detail these differences on a practical level, and how one can be generated from the other, with the hope that this will be informative to discussions regarding different SBML Level 3 Rule based formats.
Abstract: I will discuss SBML features that I believe are necessary for rule-based modeling. I will present a proposal for how to encode in SBML: 1) Multi-state multi-component species expressed as typed graphs with explicit representation of the connectivity of species’ components, 2) Reactions in the form of graph transformations, 3) Patterns that select groups of species having user-specified properties, 4) Rules that define transformations of sets of species selected by some pattern, 5) Species represented as hierarchical graphs of an arbitrary degree of nesting.
  • Will Chen (Harvard): Phospho-dynamic Experiments and Computational Modeling of the ErbB Pathways
Abstract: The ErbB receptors are a set of four plasma membrane-bound receptors that play a role in organismal development and disease, such as proliferation, invasion and metastasis of cancers. The four receptors are able to homodimerize and heterodimerize when bound to extracellular ligands, and couple to the MAP kinase and PI3K/AKT pathways to transduce extracellular signals to the nucleus. High precision, phospho-dynamic measurements are married to a biologically simple (but computationally complex) chemical reaction ODE model of early pathway events. The role of rule-based modeling for large chemical systems will be discussed in the context of studying the dynamics of the ErbB receptors.
  • Joshua Colvin
Abstract: Modeling signal transduction networks, which exhibit an enormous combinatorial complexity, is a challenging task. In the past years we have developed systematic methods to create strongly reduced models of these complex reaction networks. These methods mostly are independent of exact numerical values of the kinetic parameters. They mainly base on assumptions about domain interactions. However, the feasible interactions between domains of scaffold proteins and receptors are restricted by elementary thermodynamic constraints, which should be considered in modeling.
Abstract: Biological signalling networks can be described neatly in a language where agents bind and modify their internal states according to some specified rules. However, natural rule sets generate concurrent transition systems (or stochastic ones, given rates for rules), which are large (many agents are involved) and high dimensional (the number of non-isomorphic ways in which agents can connect is high or even infinite), so that even simple tasks such as simulation are costly. We develop here an abstract interpretation of those networks, based on the natural notion of an agent view, which leads, among other things, to an efficient computation of a superset of the system dimensions. We show this superset is the exact one, ie not just a superset, iff the latter is closed under swap (an operation whereby pairs of edges of the same type can permute their ends). That property can be taken as the definition of a local signalling network, a rather strong constraint which is implied by a simple syntactic restriction on rules. Suprisingly we can show that two substantial examples -- the EGF/FGF receptors early response pathways -- verify that syntactic condition (up to some transformations), and therefore despite their apparent complexity fall within the class of local networks.
Abstract: BioPAX is a collaborative effort to create a data exchange format for biological pathway data. Recently we have proposed rule based aspects for covering generic entities and unknown state variables for the next level of BioPAX. I will present the biological requirements driven from the data providers in the BioPAX community as well as our design rationale for the proposed features.
Abstract: Binding interactions among proteins and other biomolecules occur at the level of domains and motifs and often follow phosphorylation or other posttranslational modifications. The catalog of these functional elements and their interactions is continually growing and poses a major barrier to the development of predictive mathematical models because of combinatorial complexity, the explosion in the number of possible chemical species and reactions that can occur in such networks. The BioNetGen (BNG) language uses graphs to represent proteins and other biomolecules, with nodes representing functional subunits of these molecules and edges representing binding interactions. Graph rewriting rules describe biochemical transformations, such the formation or dissociation of bonds or state changes. This language enables the construction of precise and comprehensive models and greatly expands the scope of information that can be incorporated into models of cellular networks. BNG also incorporates a wide range of analytical and simulation tools, and networks generated by BNG can be exported in the Systems Biology Markup Language and other formats allowing interoperability with other modeling platforms. Standard methods for simulating reaction networks, such as ODEs and the Gillespie algorithm, are often not adequate to simulate networks arising from the rule-based description of realistic signaling cascades, requiring the development of simulation algorithms that avoid explicit generation of the reaction network. I will describe recent progress in the development of such algorithms and also present several applications to the modeling of signal transduction networks, including immune and growth factor receptors.
Abstract: We introduce the language CP, a compositional language for place transition petri nets for the purpose of modelling signalling pathways in complex biological systems. We give the operational semantics of the language CP by means of a proof theoretical deductive system which extends multiplicative exponential linear logic with a self-dual non-commutative logical operator. This allows to express parallel and sequential composition of processes at the same syntactic level as in process algebra, and perform logical reasoning on these processes. We demonstrate the use of the language on a model of a signaling pathway for Fc receptor-mediated phagocytosis.
Abstract: To help us understand how bioregulatory networks operate, we need a standard notation for diagrams analogous to electronic circuit diagrams. Such diagrams must surmount the difficulties posed by complex patterns of protein modifications and multiprotein complexes. To meet that challenge, we have designed the molecular interaction map (MIM) notation. MIM notation is used to build three types of diagrams: (1) explicit diagrams that define specific pathway models for computer simulation; (2) heuristic maps that organize the available information about molecular interactions and encompass the possible processes or pathways; and (3) diagrams of combinatorially complex models. Our discussion will focus on combinatorially compelx models and its applications.
Abstract: We present an agent based formalism, called the kappa-calculus [1], and a platform that enables the analysis and simulation of large-scale biomolecular reaction systems as they occur in cellular signal transduction. The framework departs from extant rule-based approaches in providing an understanding of a system's behavior directly at the level of its rule-based specification, rather than using rules as a means for generating differential equations and then using these equations to understand system dynamics. In this way, it becomes possible to pinpoint the contribution if individual rules - which describe mechanistic elements of interaction - to overall system behavior. The framework enables central concepts from concurrency theory to to be exported to biology. In particular, it formalizes the intuitive notion of "pathway", making it amenable to computational and formal analysis. The implementation of the framework includes a scalable, fully rule-based stochastic simulator whose algorithmic complexity per update cycle is independent of the number of possible molecular complexes and logarithmic in the number of rewrite rules.
[1] Formal Molecular Biology (TCS 325, 2004) Vincent Danos, Cosimo Laneve.
Abstract: I will give a general rundown of Moleculizer and its new compartmental version, Cpt, according to our common outline. I will try to give some insight into their common scheme for rule-based reaction network generation. The two programs share code for this part of their operation, though their simulation algorithms are different, their notions of "species"are different, and so forth. The template-based approach to rule expression, outlined in the proposal by Blinov et al. and related to BioNetGen's method, is very flexible and easy-to-understand. Moleculizer's approach to rule expression is peculiar, more difficult to use, and less flexible, partly because of speculative parts of reaction generation that were originally intended to dovetail with work by structural biologists at MSI. In spite of having so exquisitely abstracted Moleculizer's reaction network generation machinery, my general feeling is that its peculiarities should not be "catered to" by SBML. Rather, they invite us to consider the practicalities of using SBML with peculiar programs.
Abstract: Little b was developed to enable scientists to build mathematical models of biological systems. Users describe biological entities like molecules, reactions, compartments, cells and epithelia in a human-readable format; a rule-based inference system calculates implied reactions and species; a symbolic math system simplifies and translates mathematical expressions. These capabilities form the basis of a shareable notation which enables automated model preparation. Little b can also be viewed as a toolkit for computational biologists. The ability to extend the language and describe new approaches is central to the design philosophy. Little b is a full programming language, based in Lisp. The ability to specify functions, macros and even new syntax within the language enables programmers to develop concise notations for specific tasks.
Abstract: Biological entities may exist under different states, which are caused by covalent modifications, binding to modulators, alternate conformations or physical properties etc. Since most entities possess more than one state variable this results in a combinatorial explosion of possible states. StochSim is a mesoscopic stochastic biochemical simulator, which includes the possibility to simulate reactions between multistate complexes. The states are represented by vectors of state variables, and the number of reactions increases linearly with the number of state variables rather than exponentially. Complex configuration files are needed for setting-up a simulation. A comfortable and portable way would-be to use SBML natively. Because of the reason that SBML Level 2 does not provide structures to encode species state variables and the related reactions, this information has to be encoded in StochSim proprietary annotations. The state variables and their possible values are stored in SBML speciesTypes, the initial levels in SBML species and the description of multistate reactions in SBML reaction. We will also give an overview of how the simulator works currently and especially how the implementation of multistates is realised. Open questions remain on how best to encode instantaneous equilibriums between states and nearest-neighbour interactions in lattice. Although the SBML extension we developed is meant to be used in StochSim for now, we developed it with a much broader audience in mind, for instance not restricting us to binary state variables.
Abstract: I will be presenting the current version of SimBiology, a software package intended for the modeling, simulation and analysis of biochemical pathways. It is a MATLAB based product that exploits the flexibility of having a programming interface to all of its functionality. Hinging on this flexibility we propose possible future enhancements to add rule based modeling capabilities. The focus of the talk is to get attendees acquainted with SimBiology for the purpose of founding conversations and/or collaborations that extend rule based modeling in this environment.
Abstract: Because cellular signaling is exceedingly complex, we believe a mathematical model predicting the dynamic behavior of a system of interest is essential for understanding the system and that such a model has the potential to guide rationale drug discovery, clinical practice and diagnosis. Mechanistic models of signal transduction and other cellular systems are difficult to formulate because of the problem of combinatorial complexity, which is present whenever a relatively small number of biomolecular interactions have the potential to generate a much larger number of chemical species and reactions. For example a complete model of the ErbB signaling pathway would have require keeping track of over 100 trillion potential molecular configurations. We have recently developed a C version of network modelng software that can read BioNetGen inpiut files and has improved capabilities and speed. The software has been specifically designed to allow one to make testable in silico predictions concerning the effects of drugs individually and as combinations.
Abstract: In this talk, I will outline some recent work in both metabolic and regulatory systems that makes use of rule-based techniques to evaluate populations of competing hypotheses for the level of support or contradiction each hypothesis encounters with respect to multiple data sources. I will outline the features of some formal languages used to represent the rules and the interacting agents in several applications, and then present a possible unifying "base language" capable of expressing the salient features from all of the examples in a common framework. I will also show how these techniques can been used to proofread biomedical knowledge bases, to unify heterogenous data resources, and to evaluate hypotheses about regulation and metabolism.
Abstract: The modeling and simulation tool Simmune allows for the definition of detailed models of cell biological processes ranging from interactions between molecular binding sites to the behavior of populations of cells. Based on the inputs the user provides through a graphical interface, the software automatically constructs the resulting sets of partial differential equations describing intra- and extra-cellular reaction-diffusion and integrates them, providing numerous ways to display the behavior of the simulated systems and to interact in a way that closely resembles wet-lab manipulations with running simulations. In the talk, I will explain the technical foundations and typical use cases for simmune.
  • Ty Thomson (MIT): A Collaborative Process for Building and Documenting Biological Models
Abstract: Evaluation, reuse, and extension of published computational models is made difficult by the fact that most model builders do not provide organized descriptions of the reasoning and decisions made during model construction (e.g., choice and parameterization of reactions, system boundaries, etc.). The lack of access to such information results in individual modelers repeatedly redeveloping models, potentially using different subsets of the literature or interpretations thereof. Here we develop a principled process for describing the construction of a model along with support for model promulgation, reuse, revision, and extension. We apply our methods to develop a model for the Saccharomyces cerevisiae pheromone signal transduction pathway. The website (www.YeastPheromoneModel.org) presents both a formal rule-based model as well as the natural language documentation drawn directly from the literature. Choices and assumptions are clearly stated to facilitate their evaluation by other researchers. Modeling statements can be automatically extracted and processed into a BioNetGen model file for study and analysis. We believe that the cooperative and open development of well-documented models, enabled by web-based platforms such as wikis, would greatly increase model quality, scale, and number.

Travel Information

The Santa Fe Institute is located in Santa Fe, a small town in New Mexico, USA.

To get to the Santa Fe Institute from outside of New Mexico, you will most likely be flying into Albuquerque International Airport (code: ABQ), which is 60 miles (~100km) away from Santa Fe. It is also possible, though unlikely, that you will fly into the Santa Fe Municipal Airport (code: SAF), but few airlines make connections there. The Santa Fe Institute provides the following useful pages of information about transportation to SFI from ABQ:

The following are some suggestions and recommendations for international visitors:

  • Distances in the western USA are deceptively long (much longer than you may expect from driving in other areas of the world) and the region is sparsely populated. If you are planning on doing any sightseeing, make sure to check your distances carefully.
  • English is the dominant language in Santa Fe. Many residents also speak Spanish.
  • Airport currency exchange services are usually more expensive than taking cash out of an ATM or exchanging at a bank. You may want to check the current exchange rates (e.g., at xe.com) so that you have some idea of what to expect, although of course you will not get the exact exchange rate because all services charge a percentage for doing the exchange.
  • Tipping is customary in the USA and generally not precomputed on bills. 15%-20% is typical and expected for restaurants and taxis, tipping $2-$5 if someone carries your bags or parks your car is also typical, and leaving $2 for the hotel room cleaning staff every day is also appreciated by the staff.
  • You may wish to obtain whatever kind of power adapter is necessary for you to use your laptop and other electrical devices before you arrive into Santa Fe. It may be possible to buy converters at shops around your hotel or elsewhere in Santa Fe, but it is not guaranteed, nor is it assured they will have the right type for your electrical needs. See [3] for information about electric standards in the USA.
Retrieved from "https://wiki.santafe.edu/index.php?title=Rule-Based_Modeling_of_Biochemical_Systems&oldid=8802"