Sloppiness in ERK Phosphorylation Parameter Space

Abstract

System biology models are inherently ‘sloppy’ models where certain parameters are poorly constrained, while others can be considered ‘stiff’ and well fitted to experimental data. Consequently, there has been interest in evaluating and exploring the predictions given by possible parameter set values of concentration profiles, rather than directly determining parameter estimates from experimental data. The ERK/MEK dual phosphorylation model was studied here as an example of ‘sloppiness’ in system biology models. Through identification of key regions of interest where large numbers of test parameter sets fitted well with data generated by a reference parameter set, point clouds representing parameter spaces were determined. Calculating moments of inertia as well as applying principal component analysis to each of these volumes proved to be decent characterizations describing the space of the point clouds. These results suggest that investigation into model predictions generated by reference parameter sets may be more worthwhile and meaningful to explore than to derive parameter estimates through data fitting.