A novel systems biology framework to contextualize metabolic processes using elementary flux modes and gene expression data.

Abstract

Metabolism expresses the phenotype of living cells and understanding it is crucial for di_erent applications in biotechnology and health. However, metabolism involves an intricate network of biochemical reactions and the measure of their activity requires sophisticated methods that integrate disparate experimental data. High-throughput 'omics' technologies have posed a novel scenario where metabolism can be more globally analyzed and have brought about a new _eld of research termed metabolic systems biology. Speci_cally, organism-speci_c genome-scale metabolic networks, which de_ne the speci_c set of reactions for a given organism based on its genomic data, constitute the core of metabolic systems biology. The use of genome-scale metabolic networks is becoming more and more popular and their value has been proved for di_erent applications. In particular, these networks allow the analysis of metabolic pathways at an unprecedented level of detail and, to that end, di_erent mathematical pathway concepts have been developed. The concept of elementary _ux modes (EFMs) holds a prominent place in the _eld of metabolic systems biology, as it goes beyond prede_ned pathways and correctly captures _exibility found in metabolic systems. With the increasing availability of metabolomic, proteomic and, to a larger extent, transcriptomic data, the elucidation of speci_c metabolic properties in di_erent scenarios and cell types is a key topic in systems biology. The use of EFMs for this purpose has been limited so far, mainly because their computation has been infeasible for genome-scale metabolic networks. The purpose of this doctoral thesis is to overcome these issues and develop a novel framework for contextualizing gene expression data based on EFMs arising from genome-scale metabolic networks. Speci_cally, this thesis focuses on: The computation of a representative subset of EFMs that characterize global metabolic properties for a given cell/organism. A general statistical framework for selecting the most relevant EFMs in di_erent scenarios based on gene expression data. Application of this framework to distinguish metabolic features between various lung cancer subtypes. Dissertation structure This doctoral thesis is structured in two main parts. The _rst part contains the dissertation, including the state of the art and the main contributions made. The second part includes a complete copy of the publications that came out of the development of this thesis. In particular, the doctoral thesis is divided into the following chapters: Part I: Dissertation 1. Preliminaries. 2. EFMs computation in genome-scale metabolic networks. 3. Integration of gene expression data into EFMs. 4. Conclusions and future work. Part II: Publications Appendix A: Paper 1 Appendix B: Paper 2 Appendix C: Paper 3