GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks

arXiv:2611.04109 · q-bio.MN · Submitted 2026-11-03 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Genomics Radio. Generated commentary on the latest computational biology and genomics papers.

Ines: We're looking at the paper: "GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks".

Marcus: Here's the gist of the paper. GINsim is software for defining and dynamically analyzing qualitative models of biological regulatory networks.

Ines: This tool is significant because it introduces new functionalities for analyzing large logical models, such as symbolic computation of trap-spaces and model reversal.

Marcus: Software Capabilities GINsim is a software for the definition and dynamical analysis of qualitative (Boolean and multi-valued) models of biological regulatory networks.

Ines: It builds on its modular architecture, based on the MDDlib decision-diagram and bioLQM model-manipulation libraries.

Paper summary: Ines: So to summarize this paper, GINsim is presented as software designed for defining and dynamically analyzing qualitative models of biological regulatory networks, covering both Boolean and multi-valued systems > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg1> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Marcus: The core thesis is that version three point one adds new capabilities specifically targeting the analysis of larger logical models, focusing on the symbolic computation of trap-spaces and the Booleanisation of multilevel models > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg1> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Yuki: So, it’s not just adding a few features; it's fundamentally changing how we can handle the complexity inherent in these biological systems by giving us tools to analyze those large models more effectively > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg2> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Ines: Right. And they built this on an existing modular architecture using the MDDlib decision-diagram and bioLQM model-manipulation libraries, which shows it’s building on something solid already there > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg1> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Marcus: It's freely available under the GPL three point zero license, runs on any platform that supports Java, and it’s designed to be interoperable with other tools in the CoLoMoTo community > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg2> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Yuki: That interoperability is important because it means this isn't just a standalone tool; it fits into a larger ecosystem for studying these networks > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg2> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Ines: What matters for us is that the paper shows how you can use these new functionalities to explore the dynamical behaviors of regulatory networks in a way that is more systematic than just running simple simulations > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg1> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Marcus: It’s about moving beyond just knowing what happens in one specific case and developing a framework to analyze the potential space of behaviors across different model parameters > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg2> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Yuki: So, it gives us a way to explore those potential outcomes symbolically before we commit time to computationally expensive simulations > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg2> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Conclusion: Ines: So looking at the title, "GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks," it tells us this is about making qualitative models more rigorous for biological systems > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg1> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Marcus: And the authors, Laurent Tichit, Dominique Benielli, Olivier Chabrol, Denis Thieffry, Claudine Chaouiya, and Pedro T. Monteiro are the people building this tool > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg1> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Yuki: What does this mean for the wider field of systems biology? It suggests a new way to tackle the massive complexity that comes from studying interconnected biological processes > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg2> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Ines: It means we can use these symbolic tools to explore the whole landscape of possible behaviors for a regulatory network, not just pick one outcome and run a simulation on it > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg2> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Marcus: We’re talking about moving from a trial-and-error simulation approach to a more structured way of understanding the potential stability and oscillation patterns in these systems > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg1> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Yuki: If this works well with real biological data, it could give us much stronger predictions about how genetic changes will affect the system’s function across different conditions > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg2> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Ines: Exactly. It's about creating a framework that lets us interrogate the potential dynamics before we even get to the actual experiment > GINsim three point one: Qualitative modelling and dynamical analysis of logical regulatory networks <ref:2611.04109#pg2> <ref:2611.04109#pg1,GINsim 3.1: Qualitative modelling and dynamical analysis of logical regulatory networks>

Laurent Tichit, Dominique Benielli, Olivier Chabrol, Denis Thieffry, Claudine Chaouiya, Pedro T. Monteiro

Aix Marseille University · Institute of Biology of the Ecole Normale Sup´erieure, CNRS, INSERM · Institut Curie, MinesParis, INSERM · INESC-ID / Instituto Superior Técnico, University of Lisbon

q-bio.MN

Submitted: 2026-11-03

Updated: 2026-11-03

Comments: 22 pages, 3 tables, 5 figures. Submitted to npj systems biology

Code: https://github.com/sybila/biodivine-boolean-models

Project page: https://ginsim.github.io

License: http://creativecommons.org/licenses/by-nc-sa/4.0/

Importance score: 90/100

The gist: GINsim is software for defining and dynamically analyzing qualitative models of biological regulatory networks.

Key concepts

Logical Regulatory Graphs (LRGs)
These are directed, signed graphs representing biological networks. Vertices are proteins or genes, and arcs show regulatory interactions. Node activity is modeled using Boolean variables (0 for inactive, 1 for active) or multilevel variables with logical rules.
Trap-spaces
This tool computes the set of all reachable states within a network symbolically without running simulations. This allows researchers to understand the potential dynamic behaviors and stable states of large models efficiently.
Model Reversal
This feature generates a new model where the successors of nodes correspond to their original predecessors. This effectively reverses the direction of dynamics, helping to analyze how changes propagate backward through regulatory interactions.
Booleanisation of Model
This process converts complex multi-valued logical models into equivalent Boolean models using a specific mapping. This transformation preserves the dynamical behavior while ensuring only admissible states are explored.

Summary

GINsim is software for defining and dynamically analyzing qualitative models of biological regulatory networks. p. 0

Where the paper says it

chabrol@univ-amu. fr; Abstract GINsim is a software for the definition and dynamical analysis of qualitative (Boolean and multi-valued) models of biological regulatory networks.

This tool is significant because it introduces new functionalities for analyzing large logical models, such as symbolic computation of trap-spaces and model reversal. p. 0
Where the paper says it

Building on its modular architecture, based on the MDDlib decision-diagram and bioLQM model-manipulation libraries, GINsim version 3. 1 introduces new functionalities aimed at the analysis of large logical models: the symbolic computation of trap-spaces, the Booleanisation of multilevel models and the reversal of model dynamics.

Software Capabilities

GINsim is a software for the definition and dynamical analysis of qualitative (Boolean and multi-valued) models of biological regulatory networks. It builds on its modular architecture, based on the MDDlib decision-diagram and bioLQM model-manipulation libraries. p. 0

Where the paper says it

fr; Abstract GINsim is a software for the definition and dynamical analysis of qualitative (Boolean and multi-valued) models of biological regulatory networks. Building on its modular architecture, based on the MDDlib decision-diagram and bioLQM model-manipulation libraries, GINsim version 3.

Version 3.1 introduces new functionalities aimed at the analysis of large logical models, including symbolic computation of trap-spaces and the Booleanisation of multilevel models. p. 0
Where the paper says it

Building on its modular architecture, based on the MDDlib decision-diagram and bioLQM model-manipulation libraries, GINsim version 3. 1 introduces new functionalities aimed at the analysis of large logical models: the symbolic computation of trap-spaces, the Booleanisation of multilevel models and the reversal of model dynamics.

The software is freely available under GPL-3. p. 1
Where the paper says it

and the identification of the regulatory circuits driving multistability and oscillations. GINsim is freely available under GPL-3.

0 and runs on any platform supporting Java. p. 1
Where the paper says it

0 at https://ginsim. io/, runs on any platform supporting Java, and is interoperable with the other tools of the CoLoMoTo community.

Model Representation and Structure

Logical regulatory graphs (LRGs) are implemented in GINsim as directed, signed graphs where vertices denote regulatory proteins or genes, and arcs denote regulatory interactions. The activity of nodes is modeled by Boolean variables, taking the value 0 when inactive and 1 when active. p. 2

Where the paper says it

2 Methods Logical regulatory graph (LRG) In GINsim, Boolean or multilevel models are implemented in terms of directed, signed graphs, where vertices (also called nodes of the LRG) denote regulatory proteins or genes, while directed, signed arcs denote regulatory interactions (inhibitions, activations, or context-sensitive interactions, which can be labelled as dual or as unknown). By default, the activity of the regulatory nodes is modelled by Boolean variables, taking the value 0 when the node is considered inactive, and the value 1 when the node is considered active.

Multilevel variables can be used when biologically justified, with Boolean rules associated with each node specifying the regulatory requirement to reach a given activity level using literals such as NOT (! p. 2
Where the paper says it

However, multilevel variables can be used when biologically justified. Boolean rules are then associated with each node to specify the regulatory requirement to reach a given activity level.

), AND (&) and OR.

Dynamical Analysis Tools

GINsim provides several tools for dynamical analysis, including:

  1. Computation of state transition graphs (STG), which represents all reachable states under a chosen updating mode. p. 4

    Where the paper says it

    After defining or loading a logical regulatory graph (LRG) in GINsim, the user can analyse its dynamics by running a simulation, via the Run simulation entry of the Tools menu. This computes the STG, or a compression thereof, representing all the reachable states of the model and their admissible transitions, under a chosen updating mode (e.

  2. Computation of stable states, which uses a symbolic, simulation-free procedure to identify fixed points based on decision diagrams. p. 4

    Where the paper says it

    Computation of stable states Logical stable states typically represent persistent cellular types or phenotypes and are thus of particular interest to modellers. In this respect, GINsim implements a symbolic, simulation-free procedure that enables the exhaustive identification of all stable states (fixed points) of a logical model, based on decision diagrams [18].

  3. Identification of trap-spaces, which computes these subspaces symbolically without simulation. p. 5

    Where the paper says it

    Identification of trap-spaces Trap-spaces are subspaces of the state space that cannot be escaped by the model dynamics, independently of the updating mode; (minimal) trap-spaces provide tractable over-approximations of the attractors of a model, encompassing both stable states and complex (cyclic) attractors. GINsim computes trap-spaces symbolically, without simulation, through the Trap-spaces identification option of the Tools menu.

  4. Attractors reachability estimation using Monte Carlo and Avatar algorithms to estimate reachability probabilities from selected initial states. p. 5

    Where the paper says it

    Attractors reachability GINsim provides a dedicated tool to estimate the reachability probabilities of attractors (stable states and complex attractors) from selected sets of initial states, without requiring the exhaustive construction of the state transition graph. This analysis relies on sampling of asynchronous trajectories, either by the Monte Carlo approach or by using the Avatar algorithm [20], which is better suited to deal with complex cyclic behaviour.

  5. Model reduction functionalities that compress the regulatory graph by hiding a chosen subset of components while updating logical rules for targets. p. 6

    Where the paper says it

    To avoid this issue, one approach consists in reducing the model size. In this respect, the Reduce model entry of the Tools menu compresses a regulatory graph by hiding a chosen subset of components, while the logical rules of their targets are updated to account for the indirect effects of the hidden regulators.

Advanced Analysis Features

The software includes several advanced functionalities for deeper analysis:

- Booleanisation of model:

GINsim can convert multi-valued logical models into equivalent Boolean models using the mapping proposed by van Ham, which reproduces the dynamical behaviour while preventing the exploration of non-admissible states.

- Model reversal:

The Reverse model tool enables the generation of a model whose asynchronous successors correspond to predecessors of the original model, effectively reversing the direction of dynamics.

- Interactions functionality:

This tool computes the signs of each interaction for each regulatory context directly from Boolean rules, which is useful for identifying potential inconsistencies in multilevel nodes.

- Analyse circuits:

GINsim includes a tool to identify all or subsets of regulatory circuits and determine which are dynamically functional, helping to identify circuits contributing to multistationary or periodic behaviour.

Application Example

The paper illustrates these functionalities on a reference multilevel model of the mammalian cell cycle, including computation of its stable states and cyclic attractor. The analysis demonstrates how perturbations, such as knocking out a gene or ectopic expression, affect the reachability of attractors and the identification of functional circuits driving oscillations. p. 9

Where the paper says it

[29–31]), it turned out to be possible to reproduce the qualitative dynamical properties of systems of non-linear ordinary differential equations using the simpler Boolean framework [32–35], or moderate multilevel extensions thereof [12, 36, 37]. Hereafter, we use the multilevel model defined by Traynard et al [12] to demonstrate some of the key functionalities of GINsim, including the computation of model attractors and their visualisation, the construction and compression of dynamical graphs, the analysis of attractor reachability, and the identification of functional circuits, all of these potentially considering initial conditions, as well as simple or multiple node perturbations.

The results show that functional negative circuits account for observed oscillatory behaviour in the model. p. 6
Where the paper says it

In this respect, GINsim includes an original tool to identify all or subsets of circuits in a model, determine which of these circuits are dynamically functional (i. , contribute to the generation of multistationary or periodic behaviour) and further delineate the corresponding functionality context in terms of conditions on the values of nodes regulating the nodes involved in the circuit [18].

The paper concludes by highlighting GINsim's unique combination of a graphical interface, single archive format, multiple access modes, and support for multilevel models.

Model Repository and Integration

GINsim is distributed with a model repository gathering around seventy published logical models spanning three decades of work. It supports scripting and integration through the CoLoMoTo environment, allowing users to load, perturb, analyze, and convert models within a single reproducible workflow using Python notebooks. p. 8

Where the paper says it

From a Jupyter notebook, a model can thus be loaded, displayed, perturbed, analysed, and converted into the formats of other tools of the environment (e. , MaBoSS, Pint or pyboolnet), within a single reproducible workflow.

The paper also mentions future plans to extend the GINsim Python interface to provide access to more functionalities from the CoLoMoTo environment. p. 2
Where the paper says it

design allows the core algorithms to be reused or combined outside of the GINsim GUI. For example, from the command line, programmatically as a Java library, or from Python/Jupyter notebooks through the CoLoMoTo environment, which offers an integrated access to over twenty Boolean modelling tools [4, 15, 16].

Case Study Details

A case study on the mammalian cell cycle model shows that running the Trap-spaces identification tool with default parameters leads to two stable states when CycD is inactive and one cyclic attractor when CycD is active. A double perturbation corresponding to a knockout of Rb and ectopic activation of E2F leads to two cyclic attractors, which implies that the model predicts cyclin oscillations even in the absence of growth factors. The analysis also shows that functional negative circuits account for the observed oscillatory behaviour, while positive functional circuits presumably drive irreversible decisions along the cell cycle. p. 9

Where the paper says it

Hereafter, we use the multilevel model defined by Traynard et al [12] to demonstrate some of the key functionalities of GINsim, including the computation of model attractors and their visualisation, the construction and compression of dynamical graphs, the analysis of attractor reachability, and the identification of functional circuits, all of these potentially considering initial conditions, as well as simple or multiple node perturbations. This cell cycle model is available in the GINsim model repository at https://ginsim.

The paper emphasizes the complementarity between selected tools and resulting insights through their application to a moderate-size multilevel logical model. p. 6
Where the paper says it

To avoid this issue, one approach consists in reducing the model size. In this respect, the Reduce model entry of the Tools menu compresses a regulatory graph by hiding a chosen subset of components, while the logical rules of their targets are updated to account for the indirect effects of the hidden regulators.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Symbolic computation of trap-spaces allows for tractable over-approximations of the attractors of a model, encompassing both stable states and complex (cyclic) attractors without simulation.

  2. The system can compute stable states using a symbolic, simulation-free procedure that enables the exhaustive identification of all stable states (fixed points) of a logical model.

  3. The AI can estimate the probability of reaching attractors from initial conditions using algorithms like the Avatar algorithm, which is better suited to deal with complex cyclic behaviour.

  4. The system can identify regulatory circuits and determine which are dynamically functional (i.e., contribute to the generation of multistationary or periodic behaviour).

  5. Model reduction capabilities allow for the compression of a regulatory graph by hiding a chosen subset of components, while the logical rules of their targets are updated to account for the indirect effects.

  6. The system supports multi-valued models and can perform Model Booleanisation by converting them into equivalent Boolean models, using the mapping proposed by van Ham [25], in which each component with maximal level m is encoded by m Boolean variables.

  7. The AI can reverse model dynamics through a Reverse model tool enabling the generation of a model whose asynchronous successors correspond to the predecessors of the original model, handling multi-valued models by first booleanizing them.

  8. The system provides multiple modes of access, including a GUI, command-line and scripting interface (e.g., Jython scripts), and Python gateways based on Py4J for integration into environments like Jupyter notebooks.

Abstract

GINsim is a software for the definition and dynamical analysis of qualitative (Boolean and multi-valued) models of biological regulatory networks. Building on its modular architecture, based on the MDDlib decision-diagram and bioLQM model-manipulation libraries, GINsim version 3.1 introduces new functionalities aimed at the analysis of large logical models: the symbolic computation of trap-spaces, the Booleanisation of multilevel models and the reversal of model dynamics. GINsim 3.1 also extends the estimation of attractor reachability (Monte Carlo and Avatar algorithms), the computation of stable states, as well as model reduction functionalities. Analyses can be applied to perturbed models, and scripted, in particular from the CoLoMoTo interactive notebook. We illustrate the main functionalities of GINsim on a reference multilevel model of regulatory network that control the mammalian cell cycle, including the computation of its stable states and cyclic attractor, the assessment of their reachability, and the identification of the regulatory circuits driving multistability and oscillations. GINsim is freely available under GPL-3.0 at https://ginsim.github.io/, runs on any platform supporting Java, and is interoperable with the other tools of the CoLoMoTo community. Its website provides installation instructions and reference documentation, as well as a repository of Boolean and multilevel logical models.

Related papers