Source-linked AI summary
SBML Qualitative Models: a model representation format and infrastructure to foster interactions between qualitative modelling formalisms and tools
Claudine Chaouiya, Duncan Berenguier, Sarah M Keating, Aurelien Naldi, Martijn P. van Iersel, Nicolas Rodriguez, Andreas Dräger, Finja Büchel, Thomas Cokelaer, Bryan Kowal, Benjamin Wicks, Emanuel Gonçalves, Julien Dorier, Michel Page, Pedro T. Monteiro, Axel von Kamp, Ioannis Xenarios, Hidde de Jong, Michael Hucka, Steffen Klamt, Denis Thieffry, Nicolas Le Novère, Julio Saez-Rodriguez, Tomáš Helikar
TL;DR
Qualitative modelling lacks a standard exchange format despite its usefulness for large biological networks without precise kinetic data and the availability of many specialized tools. The paper introduces SBML qual as an SBML Level 3 extension and validates it by exchanging a signalling model among independent tools. SBML qual supports interoperability and contributes to repositories and collaborative infrastructure, while its specification remains subject to further refinement.
Problem
Qualitative models suit large networks without precise kinetic data, but existing tools lacked a standard format for exchanging models.
Method
The paper develops SBML qual as an SBML Level 3 package and tests it by analyzing a signalling network with three independent software tools.
Results
SBML qual enables exchange and interoperability among GINsim, CellNOpt, and the Cell Collective, while repositories are being prepared for SBML qual models.
Takeaways & Limitations
SBML qual facilitates combined use of complementary tools and supports collaborative development of qualitative-modeling methods and standards.
Takeaways & Limitations
The package is still being refined, including planned support for uninstantiated parameters, timing constraints, extended Petri nets, and hybrid formalisms.
Abstract
from arXiv · showhide
Background: Qualitative frameworks, especially those based on the logical discrete formalism, are increasingly used to model regulatory and signalling networks. A major advantage of these frameworks is that they do not require precise quantitative data, and that they are well-suited for studies of large networks. While numerous groups have developed specific computational tools that provide original methods to analyse qualitative models, a standard format to exchange qualitative models has been missing. Results: We present the System Biology Markup Language (SBML) Qualitative Models Package ("qual"), an extension of the SBML Level 3 standard designed for computer representation of qualitative models of biological networks. We demonstrate the interoperability of models via SBML qual through the analysis of a specific signalling network by three independent software tools. Furthermore, the cooperative development of the SBML qual format paved the way for the development of LogicalModel, an open-source model library, which will facilitate the adoption of the format as well as the collaborative development of algorithms to analyze qualitative models. Conclusion: SBML qual allows the exchange of qualitative models among a number of complementary software tools. SBML qual has the potential to promote collaborative work on the development of novel computational approaches, as well as on the specification and the analysis of comprehensive qualitative models of regulatory and signalling networks.
Background
Logical models represent biological networks with discrete component levels, logical regulatory rules, and update schemes, making them suitable for large networks lacking precise kinetic data. Before SBML qual, many specialized tools and formats existed, but no standard exchange medium adequately represented qualitative models.
- Logical modelling: Logical models assign discrete activity levels to biological components and evolve them according to logical regulatory rules and update schemes.Their dynamics can be represented as state-transition graphs whose nodes are discrete states and whose edges are state changes.
- Motivation: Logical formalisms are particularly suitable for large biological networks when precise kinetic data are unavailable.They have been applied to gene regulation, signal transduction, and cell-cycle systems across diverse organisms.
- Model construction: Logical models can be constructed bottom-up from qualitative regulatory knowledge or top-down by inferring rules from high-throughput experiments.
- Existing tools: Existing tools support complementary analyses, including model checking of attractors and paths in potentially huge state-transition graphs.Temporal logic or high-level query templates can be used with model-checking tools such as NuSMV.
- Need for standardization: A standard exchange medium was needed because qualitative models use fundamentally different building blocks from the species and reactions of core SBML.Earlier SBML representations therefore risked distorting logical models.
- Need for standardization: SBML qual was developed as a package supporting the standard definition and exchange of qualitative discrete models.Its scope also includes standard Petri Nets because of shared structural features with logical models.
- Collaborative infrastructure: The collaborative, open-source development of SBML qual helped launch CoLoMoTo, a project bringing together groups developing and using logical-modelling tools.
Development of the qual package
The SBML qual proposal was refined through community consultation and meetings among developers, then formally accepted through community and editorial approval. Supporting Java and C/C++ SBML libraries were prepared to handle the format.
- Community development: The SBML qual proposal evolved from 2008 to 2012 through community consultations and meetings involving developers of related software tools.It was accepted by community vote in 2011 and by the SBML Editors in spring 2013.
- Library support: LibSBML and JSBML provided library support for reading, writing, manipulating, and validating SBML qual content.Both libraries were available in development branches in August 2013 and planned for inclusion in major releases.
Computer simulations
The paper validates SBML qual by exchanging a signalling-network model among three tools, each offering complementary construction, simulation, reduction, or dynamical-analysis capabilities. Their analyses illustrate how update schemes, perturbations, and interactive conditions affect model behaviour.
- Interoperability validation: A signalling network was analyzed with three independent software tools to demonstrate model interoperability through SBML qual.The validation also applies complementary simulation and analysis features.
- CellNOpt: CellNOpt converts signed network graphs into compatible logic-model scaffolds, fits gates and parameters to data, and supports Boolean, fuzzy, and ODE-based formalisms.Its workflow can simplify networks by compressing intermediates and removing links involving unobservable or uncontrollable nodes.
- GINsim: GINsim supports multi-valued logical modelling, multiple update policies, state-transition-graph generation, and dynamical-property analysis.It can analyze some properties, such as stable states, without generating complete network dynamics.
- GINsim: GINsim identifies regulatory circuits and functionality contexts, while reduction methods can remove certain output species without changing attractor number, nature, or reachability.
- GINsim: Attractors capture asymptotic dynamics as stable states or terminal cyclic strongly connected components, and synchronous and asynchronous updates can produce distinct cyclic behaviours.
- Cell Collective: The Cell Collective is a web platform for constructing, simulating, and analyzing Boolean models, with sharing and SBML qual download support.Its knowledge base links model interactions to experimental research papers.
- Cell Collective: Cell Collective models can include stochastic elements and continuous input/output measures that provide semi-quantitative comparisons with laboratory experiments.
The SBML qual package
SBML qual extends SBML Level 3 with elements tailored to discrete logical models: QualitativeSpecies represent network entities, while Transitions encode rules that update outputs from inputs. FunctionTerms and defaultTerms specify the resulting discrete states.
- Package structure: SBML qual extends SBML Level 3 to support standard exchange and interoperability of discrete logical models.The format is designed around discrete state spaces and event-driven state transitions.
- Package structure: The UML representation links QualitativeSpecies to Transitions as inputs or outputs, with transitions describing how species levels change in response to other entities.
- Package structure: QualitativeSpecies represent model entities and carry discrete initialLevel and maxLevel attributes.A Boolean species may take levels 0 or 1, with maxLevel equal to 1.
- Transitions: Transitions contain Inputs, Outputs, and FunctionTerms that determine output levels from logical conditions evaluated at each time step.Inputs are referenced by the logical function, while Outputs receive values at the next step.
- Transitions: A defaultTerm supplies the result when no FunctionTerm condition applies, completing the transition rule for the participating entities.Each FunctionTerm returns a Boolean condition and specifies a resultLevel for the Outputs.
- Boolean example: A Boolean example encodes C as 1 when A=1 and B=0, with A and B as inputs and C as the output.The defaultTerm is 0, and the rule is represented with a MathML functionTerm.
Demonstration of model interoperability
SBML qual enabled one EGF/TNFα signalling model to be exported and analyzed across CellNOpt, GINsim, and Cell Collective. The tools produced consistent synchronous results while exposing additional asynchronous dynamics, perturbation effects, and simulation workflows.
- Interoperability: Three independent tools implemented SBML qual and were used to analyze the same logical signalling model.CellNOpt, GINsim, and Cell Collective provided complementary perspectives on model construction, simulation, and analysis.
- Model construction: CellNOpt derived a logical EGF/TNFα model from prior pathway knowledge and simulated experimental data, then exported it to SBML qual.The workflow recovered the model used as the example despite differences between the prior-knowledge network and the data-generating model.
- Cross-tool validation: Synchronous simulations produced consistent global evolution across tools, yielding stable states or six-state cycles depending on initial conditions.The model reached one of two stable states or one of two cycles encompassing six states.
- Dynamic analysis: Asynchronous analysis exposed much larger state-transition spaces, multiple reachable attractors, and transient oscillations that update priorities could eliminate.For EGF=1 and TNFα=1, the asynchronous graph contained about 116k states versus 19 synchronously, reducible to 546 states after removing outputs.
- Dynamic analysis: GINsim linked network circuits to dynamics: a negative IκB-NFκB-ex circuit supported oscillations, while a positive ph circuit supported attractor multiplicity.Under EGF=0 and TNFα=1, asynchronous updating could reach two cyclic attractors differing in ph; perturbations such as IKK knockout or ERK expression altered these behaviors.
- Simulation workflows: Cell Collective supported semi-quantitative input-output analyses and interactive what-if simulations of perturbations and changing external conditions.Ras gain-of-function increased Erk but not Akt, because Akt does not lie downstream of Ras.
The LogicalModel library
LogicalModel is an open-source library that supports manipulation and conversion of logical models across SBML qual and other qualitative modelling technologies. Its central interface connects formats, libraries, and software tools through import and export capabilities.
- Library functions: LogicalModel provides a data structure for manipulating logical models and supports model conversion as a standalone command-line tool.It includes SBML qual import and export through JSBML.
- Interoperability: The library connects SBML qual and logical models with additional formats and existing software tools through import and export filters.These capabilities support interoperability among qualitative modelling technologies.
- Interoperability: Figure 7 depicts LogicalModel as an interface linking formats, software tools, and libraries, with arrows indicating export and import capabilities.The ecosystem includes tools and importers for generating qualitative models from pathway databases.
Conclusions and prospects
SBML qual provides a standard exchange format for logical models and supports interoperability across complementary qualitative-modelling tools. The paper also outlines repositories, reproducibility standards, hybrid-model extensions, and community-driven refinement as future directions.
- SBML qual provides a standard exchange medium for logical models of regulatory and signalling networks, extending modular SBML Level 3 capabilities.
- GINsim, CellNOpt, and Cell Collective implement SBML qual, while additional tools can export models to the format.
- Simulations and analyses of an EGF/TNFα Boolean signalling model demonstrated consistency across the three implementing tools.
- Repositories such as Cell Collective and BioModels are preparing or hosting downloadable SBML qual models.
- Planned developments include community refinement, hybrid-formalism support, standardized simulation descriptions under MIASE, and integration with SBML Core concepts.
- SBML qual and the CoLoMoTo consortium are intended to foster collaborative development of standards and computational methods for qualitative biological-network modelling.
Supplemental Information
The supplemental simulations show that the example network reaches either a steady state or a six-state cyclical attractor depending on the EGF and TNFα inputs.
- With EGF=TNFα=0, the network reaches a steady state after 3 transient states.
- With EGF=1 and TNFα=0, the network reaches a steady state after 14 transient states.
- With EGF=0 and TNFα=1, the network reaches a 6-cycle attractor after 5 transient states.
- With EGF=TNFα=1, the network reaches a six-state cyclical attractor after 12 transient states.