17/04/2012 - Room C209/10 at 14:00 - External Seminar
Carito Guziolowski
(University Hospital Heidelberg)
The Sign Consistency Model (Siegel et al., 2006) represents a regulatory network and experimental measurements as a system of discrete qualitative constraints. Thus, the combinatorial complexity of reasoning over a large-scale regulatory network can be approached by using ecient solvers over such a system of constraints. Currently, two computational frameworks, based on the Sign Consistency Model, exist: BioQuali (Guziolowski et al., 2009) and BioASP (Gebser et al., 2010). Both address and implement computationally complex combinatorial questions that appear when confronting discrete large-scale regulatory networks with OMIC experimental data. In this talk I will focus on two of our recent results using these frameworks. The rst refers to a method based on Answer Set Programming (ASP) that prunes an initial (generic) network topology in order to consider only those interactions that explain experimental outputs. The signaling network and phosphoproteomics data are in-silico simulated from the network used in the DREAM Predictive Signaling Network Challenge (Prill et al., 2011). Firstly, we compared our results with those obtained by Cell Net Optimizer (Saez-Rodriguez et al., 2009). Secondly, we assessed the impact over the optimization process caused by the number of measurements and the network size. These results, hard to be interpreted intuitively, can be straightforwardly generated with ASP, since it provides complete solutions for the optimization problem with extremely fast computation times. The perspective of this research goes towards proposing automatically optimal experimental designs that will improve the optimization of generic networks. The second result is a method that automatically extracts curated regulatory information from public databases, trans- forms this information into computable logical networks, reduces the graph topology to only relevant information by using microarray data, and, nally, confronts this data logically to the network topology using BioQuali. The outputs of our method are networks that are mechanistically linking growth factor receptors, intracellular signaling, and e ective cell state changes. The modeling of such networks is essential for understanding how cell-cell communication controls cellular behavior in the end. This last work addresses the general problem that the large amount of transcriptomic data currently available strongly contrasts with the few discrete and qualitative representations of biological models that can be in-silico analyzed. References Gebser, M., Ko andnig, A., Schaub, T., Thiele, S., and Veber, P. (2010). The bioasp library: Asp solutions for systems biology. In Tools with Arti cial Intelligence (ICTAI), 2010 22nd IEEE International Conference on, volume 1, pages 383 {389. Guziolowski, C., Bourde, A., Moreews, F., and Siegel, A. (2009). BioQuali Cytoscape plugin: analysing the global consistency of regulatory networks. BMC Genomics, 10, 244. Prill, R. J., Saez-Rodriguez, J., Alexopoulos, L. G., Sorger, P. K., and Stolovitzky, G. (2011). Crowdsourcing Network Inference: The DREAM Predictive Signaling Network Challenge. Sci Signal, 4, mr7. Saez-Rodriguez, J., Alexopoulos, L. G., Epperlein, J., Samaga, R., Lau enburger, D. A., Klamt, S., and Sorger, P. K. (2009). Discrete logic modelling as a means to link protein signalling networks with functional analysis of mammalian signal transduction. Mol. Syst. Biol., 5, 331. Siegel, A., Radulescu, O., Le Borgne, M., Veber, P., Ouy, J., and Lagarrigue, S. (2006). Qualitative analysis of the relation between DNA microarray data and behavioral models of regulation networks. BioSystems, 84, 153{174.