WebMaBoSS: A Web Interface for Simulating Boolean Models Stochastically

WebMaBoSS is an easy-to-use web interface for conversion, storage, simulation and analysis of Boolean models that allows to get insight from these models without any specific knowledge of modeling or coding. It relies on an existing software, MaBoSS, which simulates Boolean models using a stochastic approach: it applies continuous time Markov processes over the Boolean network. It was initially built to fill the gap between Boolean and continuous formalisms, i.e., providing semi-quantitative results using a simple representation with a minimum number of parameters to fit. The goal of WebMaBoSS is to simplify the use and the analysis of Boolean models coping with two main issues: 1) the simulation of Boolean models of intracellular processes with MaBoSS, or any modeling tool, may appear as non-intuitive for non-experts; 2) the simulation of already-published models available in current model databases (e.g., Cell Collective, BioModels) may require some extra steps to ensure compatibility with modeling tools such as MaBoSS. With WebMaBoSS, new models can be created or imported directly from existing databases. They can then be simulated, modified and stored in personal folders. Model simulations are performed easily, results visualized interactively, and figures can be exported in a preferred format. Extensive model analyses such as mutant screening or parameter sensitivity can also be performed. For all these tasks, results are stored and can be subsequently filtered to look for specific outputs. This web interface can be accessed at the address: https://maboss.curie.fr/webmaboss/ and deployed locally using docker. This application is open-source under LGPL license, and available at https://github.com/sysbio-curie/WebMaBoSS.

When you create a project, you have two options. You can: -Load a model from your own desktop: in this case, you can choose between four predefined formats: MaBoSS, SBML-Qual, GINsim or BoolNet (BNet). -Import a model from BioModels or Cell Collective: the server is linked to the content of the two databases and the models can be imported directly. In the future, we can extend For this tutorial we study a model of T-helper cells for regulation of IL-17A/IL-17F from Corral-Jara, K.F., Chauvin, C., Abou-Jaoudé, W. et al. imported from BioModels.
Once imported, the user can change the name of the model, export it in a new format or delete it. Clicking on the name of the model allows the user to edit and elaborate the model itself.

❖ View the content of the model o Overview Tab
Once imported, the model can be visualized. Note that the layout is conserved if the model is imported from GINsim where the layout is encoded in the model description.
This model is composed of 82 nodes and 136 links, describing the interplay between different transcription factors involved in the regulation of IL-17A and Il-17F.
From this model analysis, it was concluded that NFAT2A, STAT5A and SMAD2 are key regulators of the differential expression of IL-17A and IL-17F.
o Editing tab Models exported in MaBoSS are composed of two files with the extensions .bnd and .cfg: a bnd file for the model description (logical rules) and a cfg file for the setting of the model and simulation parameters.
With WebMaBoSS, the .cfg file can be accessed through the editing tab. The user can visualize the logical equations for each node. It is possible to: (1) edit the formula using the buttons on the right of the corresponding nodes or (2) insert a new node in the network by clicking on the "New node" button at the bottom of the list of nodes.
It is possible to set also modify the content of the .cfg file by setting different initial conditions, different outputs (variables shown explicitly in the simulations), transitions rates and simulation parameters.
o Simulation tab MaBoSS computes trajectories over time and the mean probability of the model states of interest. When launching a new simulation, one can name it, set the maximum time length, select the number of trajectories.
The initial conditions need to be set for each node. They can be all random or all set to 0 depending on the format of the files imported.
• Th1 condition, with initial state: IL12_In • Th17 condition with initial state: IL1_In, IL23_In, TGFB_In, IL6_In • IL-12 + IL-1β condition, with initial state: IL1_In, IL12_In A subset of nodes is set to 1 (referred to as "common nodes"). On top of this definition of the "common" initial condition, these three configurations are set to account for different cell conditions. They can be found in the MaBoSS repository as .cfg file (see Corral_tutorial.cfg).
Another important point is to select the proper outputs for the simulations. MaBoSS uses all variables for computing the probabilities but if all variables are kept for the output, we may face a computational explosion. We suggest to choose only a subset of variables and launch several simulations with different outputs rather than keep all the variables of interest in the same simulation. The results can be visualized as a pie chart (final state distribution), node probability trajectories, state probability trajectories, and fixed points.
▫ Nodes probability trajectories plot: ▫ It is possible to visualize directly on the network the active nodes in the selected stable state, using the blue eye icon on the left.
The Corral model proposes a combinatorial roles of SMAD2, NFAT2 and STAT5A in the differential expression of IL-17A and IL-17F. According to the table in fig. 6 of the initial publication, we can introduce a perturbation to check the different levels of expression of IL-17A based on the knock-in or knock-out of SMAD2 and STAT5A.
To introduce a perturbation in the model with WebMaboSS, the user can select one or more nodes in the Mutations tab and run a new simulation, as shown in the figure below: The perturbation can represent a gene overexpression (green switch) or a knock-out (red switch).
In the case shown above, we simulated with IL-12 + IL-1b condition, a knock-out of SMAD2 and a knock-in of STAT5A. This mutant leads to a higher activity of IL-17F and an inhibition of the activity of IL-17A, which, in the absence of SMAD2, is caused by STAT5A, itself activating IL-17F.
It is easy to verify the effect of the perturbation by looking at the States probability trajectories graph: IL-17F is expressed along with IL2 while there is no expression of IL-17A.
❖ Sensitivity analysis With WebMaBoSS, it is possible to perform a sensitivity analysis to check the effect of a knock-in/knock-out of each node on the network.
In the General setting tab, the user should decide whether to introduce a single or a double mutation and if this one would be a knockout (OFF), an overexpression (ON) or both. WebMaBoSS will simulate a perturbation on the model one node at a time (or two if the user select double mutation). Depending on the size of the network, this analysis will be quite expansive in terms of time, so, to avoid longer computational time, in the Candidates tab, the user should select the interested nodes to mutate. Finally, as seen previously, in the Output tab, the user should select a subset of nodes to visualize to avoid a computational crash.
As example, we performed a sensitivity analysis of the model, choosing as candidates for the mutation SMAD2, STAT5A_b1, STAT5A_b2, NFAT2A_b1, NFAT2A_b2 and as output IL2, IL17A and IL17F.
At first, WebMaBoSS will print all the possible results, but using the filter function provided by WebMaBoSS, we can verify which mutation leads to a certain phenotype.
1) We want to look which mutations leads to a IL17A phenotype. We select the state we want to search for (IL17A) and a superior/inferior threshold (superior, 0.30). WebMaBoSS filters all the results until it finds the ones corresponding to the setting.
2) This time, we are searching for IL17F phenotype. We select in the filter the corresponding state (IL17F--IL2) and a threshold superior than 0.35: