Two characteristic properties of biochemical (in particular signaling) networks are their complexity and their inherent modularity. The main idea of my Ph.D. work was to take advantage of the latter to facilitate to unravel the former. To apply such a modular approach, we worked on the following tasks:

1. Decompose the signaling network of interest into sensible modules,

2. Analyze the modules thoroughly, and

3. Rewire the modules (eventually substituted by a simplified version) into the network and analyze the network.

We defined a novel criterion for the decomposition of networks based on the concept of retroactivity (Saez-Rodriguez et al. IEEE CSM 2004, Saez-Rodriguez et al. Comp. Chem. Eng. 2005), and we developed an algorithm to define modules so that the number of retroactive interconnections is minimized, thus obtaining subsystems as uncoupled as possible (Saez-Rodriguez, Gayer, et al., Bioinformatics 2008).  Retroactivity, the effect of downstream elements on the state of the upstream element, is an important issue in systems theory, and also significant in biochemical systems: retroactive connections modify the functioning of a module, which is then dependent on what is downstream of it, complicating the analysis.

We then analyzed from a system-theoretical point of view a set of modules describing fundamental units of signaling networks (corresponding to individual proteins). We found that their properties are very simple (Saez-Rodriguez et al., IET Syst. Biol. 2008). This suggests an analysis based on describing them using relatively simple Boolean (logical) operators, which can then be wired together to represent large networks.

Based on this simplification, we developed a new analytic framework (Klamt, Saez-Rodriguez, et al., BMC Bioinformatics, 2006) and the software tools necessary to set up (ProMoT, Saez-Rodriguez, Mirschel, et al., BMC Bioinformatics, 2007) and analyze ( CellNetAnalyzer , Klamt et al., BMC Syst. Biology, 2007) Boolean models of signaling circuits. To test their utility, we constructed a model comprising 94 proteins describing the signaling network responsible for the activation of T-cells, key players of the immune system. The model predicted several unexpected features of the signaling system that were subsequently validated experimentally (Saez-Rodriguez et al., PLoS Comput. Biology, 2007). 

You can find more about my Ph.D. work here.