Sammendrag
Metabolism is the set of biochemical reactions that occur within a living organism in or- der to maintain life and grow. Most of these reactions are catalysed by enzymes which are coded for by genes. Using existing biochemical, genetic and genomic knowledge, one can link reactions together into pathways and further into metabolic networks, accounting for all enzyme-coding genes and which reactions they catalyse. Thus, metabolic networks can be made for entire organisms from their sequenced and annotated genome. As a means to predict network functionality and phenotypes, they are converted into genome-scale metabolic models (GEMs). GEMs are increasingly used to study the physiology of vari- ous organisms, ranging from microbes to complex multicellular eukaryotes, in order to understand and possibly benefit from their metabolic activities. However, there is increas- ing evidence that only a subset of metabolic reactions in a network is active in any given context, making GEMs superfluous when specific conditions are investigated. Therefore, several methods have been developed to extract context-specific metabolic models by in- tegrating omics data with GEMs. Although context-specific models are assumed to yield more accurate predictions of phenotypes in a particular context, their accuracy regard- ing metabolic functionality has not yet been sufficiently tested. To overcome this, I here assess the capability of six model extraction methods (MEMs) to create functionally ac- curate context-specific models, using an Atlantic salmon GEM and hepatic transcriptomic data. To this end, I extend current methods for predicting sample-specific activity states of metabolic tasks to overcome the particular challenge of not having an objective truth to benchmark against in MEM comparisons. Context-specific models outperformed the GEM from which they were built, indicating that context-specific modelling captures real- istic representations of metabolism in a given context and thus yield practical and biologic- ally meaningful predictions. These results support current evidence that context-specific models are advantageous when studying the metabolic behaviour of organisms, especially when investigating specific contexts of interest. The findings of this study contribute to the current knowledge regarding context-specific metabolic modelling and may facilitate further research. Consequently, this may potentially be beneficial for both academic and industrial purposes.
Vis fullstendig beskrivelse
