Fig. 5

Analysing the robustness of DMPy. Using prior information for kinetic rates in the L. lactis system, a model is created whereby the distribution of all parameters is known. From these simulations we select a subset of the data to act as a ‘gold standard’ set of experimentally measured time-series of system components. We then use a subset of the known parameter distributions as an input to create a new model of L. lactis metabolism, whose dynamics are compared against the ‘gold standard’ dataset. Given a percentage of known parameter values used as input into the pipeline, the posterior distribution is resampled until the mean error score between model simulations and the ‘gold standard’ dataset converges (Fig. 4). The mean error score is recorded for this subset of parameters, after which a new subset is chosen and the process is repeated (n=10000). Only successfully converged subsets are shown. See Additional file 1: Figure S2 for more information