Abstract for the talk at 17.12.2013 (15:15 h)VW Seminar
Jakob Lykke Andersen (University of South Denmark)
Transformation Patterns in Chemical Reaction Networks
Several established methods for pathway analysis in chemical reaction networks, e.g., Flux Balance Analysis and Extreme Pathways, use the notion of flux vectors, defined through linear algebra and the representation of the networks as stoichiometric matrices. Fluxes are however vectors of real numbers even though the number of times a reaction happens can only be integer. Some reaction networks can additionally not even be modelled by stoichiometric matrices without loss of information. We present a new model based on a generalisation of classical network flows to flows on directed multi-hypergraphs, which in a relaxed configuration model fluxes, but with integer constraints properly models chemical pathways for all classes of reaction networks. The model is augmented with constraints for finding specialised transformation patterns such as autocatalytic pathways, which are only well-defined for integer flows. Feasible solutions are found using integer linear programming and a custom tree search. We will also briefly present a graph grammar based approach for generating reaction networks, which coupled with the pathway model provides a powerful framework for analysing artificial chemistries.