Automated inference of Boolean models from molecular interaction maps using CaSQ

Abstract

Motivation: Molecular interaction maps have emerged as a meaningful way of representing biological mechanisms in a comprehensive and systematic manner. However, their static nature provides limited insights to the emerging behaviour of the described biological system under different conditions. Computational modelling provides the means to study dynamic properties through in silico simulations and perturbations. We aim to bridge the gap between static and dynamic representations of biological systems with CaSQ, a software tool that infers Boolean rules based on the topology and semantics of molecular interaction maps built with CellDesigner.

Results: We developed CaSQ by defining conversion rules and logical formulas for inferred Boolean models according to the topology and the annotations of the starting molecular interaction maps. We used CaSQ to produce executable files of existing molecular maps that differ in size, complexity and the use of Systems Biology Graphical Notation (SBGN) standards. We also compared, where possible, the manually built logical models corresponding to a molecular map to the ones inferred by CaSQ. The tool is able to process large and complex maps built with CellDesigner (either following SBGN standards or not) and produce Boolean models in a standard output format, Systems Biology Marked Up Language-qualitative (SBML-qual), that can be further analyzed using popular modelling tools. References, annotations and layout of the CellDesigner molecular map are retained in the obtained model, facilitating interoperability and model reusability.

Availability and implementation: The present tool is available online: https://lifeware.inria.fr/∼soliman/post/casq/ and distributed as a Python package under the GNU GPLv3 license. The code can be accessed here: https://gitlab.inria.fr/soliman/casq.

Supplementary information: Supplementary data are available at Bioinformatics online.

Fig. 1.

The repertoire of CellDesigner graphical notation schemes used to illustrate CaSQ’s rules. For CaSQ’s conversion rules, we use the notation schemes for association, transport, catalysis, state transition and also the glyphs for receptor, protein, modified protein (here, we show phosphorylation as an example) and the empty set. The empty set can account for degradation or in SBGN-PD terms, can represent the creation (respectively, the disappearance) of an entity from an unspecified source (resp. sink) that we do not need or wish to explicit

Fig. 2.

Illustration of the 1st rule. If two species of the map are only reactants in a heterodimer association, and if one of the reactants is annotated as a receptor, then the receptor is deleted from the map (its annotations are added to the product of the reaction)

Fig. 3.

Illustration of the 2nd rule. Compression of the complex formation, where none of the reactants is denoted as a receptor, and both reactants do not participate in any other reaction. As a result, both reactants are removed and modifiers are rewired to have the complex as a product

Fig. 4.

Illustration of the 3rd rule. Removing inactive forms that do not participate in other reactions

Fig. 5.

Combination of rules 2 and 3. CaSQ retains components that contribute further to the propagation of the signal

Fig. 6.

Combination of the 2nd and the 4th rule. Components that are translocated across other compartments (e.g. transcription factors) are merged in one component that inherits all influences, provided that the original component does not participate in another reaction/regulation

Fig. 7.

(a) Screenshot of simulations for Btk knockout of the CaSQ-derived mast cell activation model using Cell Collective. When Btk is set to zero, Erk and PLCG1 are not expressed. (b) Screenshot of simulations for Syk knockout of the CaSQ-derived mast cell activation model using Cell Collective. When Syk is set to zero, Erk, JNK, NFAT, NFkB, Ca2+, PKC, Elk1, PLCG1 are not expressed

Fig. 8.

Simulations of the CaSQ-inferred model using the modelling platform Cell Collective. The CaSQ-inferred model for MAPK was able to reproduce known biological scenarios, either completely or partially. The results of the in silico simulations for the three first biological conditions described in Table 3 showed perfect agreement with the results of manually built model, as depicted in a, b and c. For conditions described in scenarios 4 and 5 of Table 3, the CaSQ-inferred model could partially reproduce the attended behaviour (d and e) while simulation results for scenario 6, were inconsistent with the literature and the results of the manually built model (f, g and h)