System biology aims at understanding and analyzing biological systems using mathematical and computational models. The explosion of the number of experimental data leads to larger and larger models. In order to be able to easily analyze them and quickly simulate them, it is necessary to be able to simplify them. In this thesis, we propose simplification methods for biochemical reaction networks. These methods are sufficiently rich to be able to simplify an important number of networks from biological applications. They are contextual, allowing to consider a network as a sub-module of a larger model, and to simplify it without modifying the behavior of the global model. Finally, our simplifications are sound, meaning that they preserve the semantics of the networks. Firstly, we are interested in a non deterministic semantics, based on the capability to converge to some attractors of the network. Then we study some simplifications for the deterministic semantics, allowing for instance to remove intermediate species at steady-state. Finally, we are interested by the confluence of this simplification, as well as the relation between the elimination of intermediate species and the computation of the elementary modes of a network.
- Directeurs de thèse : LHOUSSAINE Cédric, NIEHREN Joachim - Rapporteurs : FAGES François, HAAR Stefan - Examinateurs : ROPERS Delphine, TOUZET Hélène
Thesis of the team BioComputing defended on 28/02/2017