Reaction-contingency based bipartite Boolean modelling

Abstract

Background: Intracellular signalling systems are highly complex, rendering mathematical modelling of large signalling networks infeasible or impractical. Boolean modelling provides one feasible approach to whole-network modelling, but at the cost of dequantification and decontextualisation of activation. That is, these models cannot distinguish between different downstream roles played by the same component activated in different contexts. Results: Here, we address this with a bipartite Boolean modelling approach. Briefly, we use a state oriented approach with separate update rules based on reactions and contingencies. This approach retains contextual activation information and distinguishes distinct signals passing through a single component. Furthermore, we integrate this approach in the rxncon framework to support automatic model generation and iterative model definition and validation. We benchmark this method with the previously mapped MAP kinase network in yeast, showing that minor adjustments suffice to produce a functional network description. Conclusions: Taken together, we (i) present a bipartite Boolean modelling approach that retains contextual activation information, (ii) provide software support for automatic model generation, visualisation and simulation, and (iii) demonstrate its use for iterative model generation and validation.

A scalable method for parameter-free simulation and validation of mechanistic cellular signal transduction network models

“…The bipartite Boolean modelling formalism we presented here if fundamentally different from its previous incarnation 14 . First, it has been designed to capture system level behaviour, not the states of individual molecule instances in the network, meaning that states that would be mutually exclusive on a single molecule can be true at the same time.…”

“…Fourth, the reimplementation has improved the model generation, enabled the use of different export options, and improved the model creation and analysis workflow. While the previous incarnation worked well in many instances 14,15,26 , these models had issues with certain reaction types (most notably degradation) and spurious oscillations. That the latter appeared so rare was due to the implicit dominance of modified states: Neutral states were not explicitly represented, making modification or binding reactions dominant over reactions that returned components to their neutral state.…”

“…We chose a simplified model of the HOG MAP kinase pathway from Saccharomyces cerevisiae (Fig 2; taken from 14 ). We created a rxncon 2.0 model of this pathway (Table S1), and used this to generate the bBM using the generic update rules with smoothing.…”

“…These can all be used for detailed description of signalling events. However, the first is based on microstate description, inheriting the scalability issues of these approaches 12 ; the second requires a full parametrised rulebased model, inheriting the problem of parametrisation 13 ; and the third inherited shortcomings in expressiveness and precision from the first generation of the rxncon language 14 . However, the bBM method was successfully used for validation and gap-filling of large signal transduction models 14,15 .…”

“…they describe functionality at the component level, not at the level of states, which is insufficient to describe the often context dependent activity of signal transduction components. To our knowledge, only three methods have been developed for mechanistic Boolean simulation of signal transduction networks: One derived from SBGN-PD diagrams, one from rule-based models, and one bipartite Boolean modelling (bBM) method based on the rxncon language [12][13][14] . These can all be used for detailed description of signalling events.…”