Prediction of dynamical properties of biochemical pathways with graph neural networks

Biochemical pathways are often represented as graphs, in which nodes and edges give a qualitative description of the modeled reactions, while node and edge labels provide quantitative details such as kinetic and stoichiometric parameters. Dynamical properties of biochemical pathways are usually assessed by performing numerical (ODE-based) or stochastic simulations in which quantitative parameters are essential. These simulation methods are often computationally very expensive, in particular when property assessment requires varying parameters such as initial concentrations of molecules. In this paper we propose the use of a Deep Neural Network (DNN) to predict such dynamical properties relying only on the graph structure. In particular, our model is based on Graph Neural Networks. We focus on the dynamical property of concentration robustness, which is the ability of the pathway to maintain the concentration of some molecules within certain intervals despite of perturbation in the initial concentration of other molecules. The use of DNNs can allow robustness to be predicted by avoiding the burden of performing a huge number of numerical or stochastic simulations. Moreover, once trained, the model could be applied to predicting robustness properties for pathways in which quantitative parameters are not available.