Return to Quantitative Model Refinement

Refinement of the EGFR signaling pathway

The EGFR signalling network is an evolutionary developmental pathway, which regulates various physiological responses of the mammalian cell, such as growth, survival, proliferation, differentiation and motility and plays a major role in oncogenesis, see [1, 2]. Anomalies found in the EGFR signalling pathway have been associated with various cancer types and the pharmaceutical intervention developed to tackle these abnormalities proved to be successful in the treatment of some cancer types, see [1,3].

We focus on a model of the EGF-induced signal transduction through the MAPK cascade introduced in [4], consisting of 148 reactions and 103 reactants. The model comprises a negative feedback loop from the dual phosphorylated ERK (ERK-PP) to the Sos protein, which brings about the unbinding of Grb2-Sos from the receptor complex. The model does not take into consideration protein isoform specificity (several forms of the same protein), see [12]. In the absence of the stimulus EGF, the system depicted by the model in [4] exhibits a stable steady state, which corresponds to a state of unphosphorylated ERK. The model draws a distinction between two pools of double-phosphorylated ERK, one of which is situated in the cytoplasm and one correlated with the internalization process.

Chen et al [5] constructed an expanded model of the EGFR signaling pathway by considering all the receptors of the ErbB family: ErbB1 (EGFR), ErbB2 (HER2), ErbB3, ErbB4. The authors reported extensive effort for building and fitting the model. In particular, out of about 2000 independent parameter optimization runs searching across 106 parameter sets, only about 100 sets were found to have a good fit to available data.

In our approach we relied on fit-preserving refinement [6,7] to systematically generate a model that is on the same level of detail as that of Chen et all starting from the model introduced in [12]. Moreover, based on [7], the parameters were set in such a way that our refined model follows the same dynamics with respect to the species of the original model and, thus, already fits the same data. Generating large models via fit-preserving refinement can allow for the exploration of the parameter space starting from values that already provide reasonable fits. Furthermore, the effort required for constructing the model with this approach is significantly smaller than constructing the model from scratch.

Our Copasi-based implementation of the initial EGFR signalling pathway [12] can be found here. The final model that we have obtained vie fit-preserving refinement is here. Event-B model for the basic ErbB signalling pathway can be found here and their pdf printouts can be found here.


[1] Marc R Birtwistle, Mariko Hatakeyama, Noriko Yumoto, Babatunde A Ogunnaike, Jan B Hoek, and Boris N Kholodenko. Ligand-dependent responses of the ErbB signaling network: experimental and modeling analyses. Molecular systems biology, 3(1), 2007.

[2] Kanae Oda, Yukiko Matsuoka, Akira Funahashi, and Hiroaki Kitano. A comprehensive pathway map of epidermal growth factor receptor signaling. Molecular systems biology, 1(1), 2005.

[3] Yosef Yarden and Mark X Sliwkowski. Untangling the ErbB signalling network. Nature reviews Molecular cell biology, 2(2):127–137, 2001.

[4] Jorrit J Hornberg, Bernd Binder, Frank J Bruggeman, Birgit Schoeberl, Reinhart Heinrich, and Hans V Westerhoff. Control of MAPK signalling: from complexity to what really matters. Oncogene, 24(36):5533–5542, 2005.

[5] William W Chen, Birgit Schoeberl, Paul J Jasper, Mario Niepel, Ulrik B Nielsen, Douglas A Lauffenburger, and Peter K Sorger. Input–output behavior of ErbB signaling pathways as revealed by a mass action model trained against dynamic data. Molecular systems biology, 5(1), 2009.

[6] Bogdan Iancu, Elena Czeizler, Eugen Czeizler, and Ion Petre. Quantitative refinement of reaction models. IJUC, 8(5-6):529–550, 2012.

[7] Cristian Gratie and Ion Petre. Fit-preserving data refinement of mass-action reaction networks. In Language, Life, Limits, pages 204–213. Springer, 2014.