When modelling subcellular signalling pathways, experimental data are integrated into a precise and structured framework from which it is possible to make predictions that could be tested experimentally, thereby facilitating the understanding of the biological mechanisms involved. The quantitative experimental data that are used for building the models are often sparse as compared to the size and complexity of the modelled system, and the translation of these data into dynamical models therefore leaves numerous uncertainties in parameter values. An explicit description of this uncertainty is useful in order to precisely describe assumptions that are made during the modelling process concerning parameters as well as the data that used for parameter estimation.
We have earlier developed a workflow for building and testing intracellular signalling models and for the quantification of model parameter uncertainty, and its propagation to predictions [1]. This workflow was applied on a model describing calcium (Ca)-dependent activation of Calmodulin (CaM), Protein phosphatase 2B (PP2B) and Ca/CaM-dependent protein kinase II (CaMKII) [2]. We here develop this workflow further and also apply it on the G-protein coupled cascade underlying endocannabinoid production [3]. Both CaMKII and endocannabinoids (eCBs) are important for synaptic plasticity in many brain areas. While CaMKII is often involved in LTP, eCBs rather promote LTD.
The model parameter uncertainty analysis is performed through a Bayesian sampling of the region of parameter values which produced a good fit to available experimental data. Two different methods are used: data-set iterative Approximate Bayesian Computation (ABC) [4,1] and Simplified Manifold Metropolis Adjusted Langevin Algorithm (SMMALA) [5]. This estimation of posterior parameter uncertainty translates to uncertainties in predictions made from the model, and finally, a global sensitivity analysis helps in determining the importance and role of different parameters for different model outputs.