Parameter estimation and model selection results for NGF-induced Erk1/2 signalling (pathway model A).

For each biological hypothesis (H1: no subpopulations; H2: different levels of total Erk, ; H3: different levels of total TrkA, ) three models, differing in the distribution assumption (normal vs. log-normal) and the ODE constrained properties (subpopulation mean vs. subpopulation median), have been specified and fitted to the experimental data using multi-start local optimisation (accuracy: 10 digits). For the model selection using the Bayesian information criterion (BIC) we found that H1 and H2 could be rejected according to [85] as . The subpopulations seemed to follow a log-normal distribution. The two models, and , which only differ in the ODE constrained property were acceptable.