Propuestas y análisis de técnicas para la generación de subconjuntos significativos de soluciones de redes metabólicas

Zusammenfassung

Abstract The study of metabolic networks applied to Biotechnology and to the investigation of diseases is a topic of current investigation in our days. In this sense, the availability of increasingly precise biological models makes possible the use of mathematical and computer tools for this study, which allows, in particular, to analyze the possible metabolic states of these networks. It is well known that the set of possible states of a metabolic network is an infinite set, but that it can be studied through a finite subset of states (the so-called elementary modes or EFMs of the network) that allow us to represent all the other possible states depending on them. Because of this, different extraction methods have been proposed for EFMs based on different mathematical strategies and tools. Each proposed method has its own advantages and disadvantages in terms of efficiency, scalability, etc. If we look at the term efficiency, the tool that has shown the best performance to date is optimization with linear programming, although it also has its limitations. It should be noted that, despite all efforts and regardless of the method used, the problem of finding the set of all EFMs is still open when it comes to large networks, so it is necessary to focus on finding subsets of EFMs, trying to ensure the biological representativeness of the subset obtained. This thesis contributes to the study of these topics based on the analysis of the different algorithms that can be proposed. It culminates in the proposal of the EFM-Ta algorithm for the extraction of EFMs, which breaks the barrier of the ideal ratio of efficiency to LP, beating in this respect previous techniques. In addition to this algorithm that dramatically improves the efficiency of previous methods, it is worth noting the inclusion of a statistical study technique on the typology of the EFMs obtained through different extraction methods. We consider that this analysis is a step forward for a study that will allow to shed light on the representativeness of different subsets of EFMs.