| Genome-scale metabolic models (GEMs) include the genome wide information of metabolic reactions and their enzymatic and genetic associations in a given organism or cell. They have been commonly used in studies of metabolic engineering in the past decade, and recently, many attempts were made to apply them in studies of other fields such as biotechnology and systems medicine. However, there are several general problems in applications of GEMs. For instances, the computational cost in simulations are high in applications of metabolic engineering, the gene-protein-reaction associations in GEMs are oversimplified, and there’s no transcriptional regulatory or signal transductional included in GEMs. These problems impeded the wider use of GEMs.A novel in silico metabolic engineering method, namely IdealKnock, was developed in this study. This method greatly improved the efficiency of prediction of high level reaction knockout metabolic engineering strategies based on GEMs. IdealKnock could identify knockout strategies within several minutes, and increase the number of knockout won’t exponentially increase the computational cost. In addition, the resulting mutants identified by IdealKnock exhibited superior outcome in both growth and production compared to other methods. Moveover, through the visualization, the principles behind several selected knockout strategies were interpreted, which were coupling growth and production, and blocking competing pathways. The interpretation of these strategies could be helpful in further engineering of the strains.In this study, logical transformation of GEMs (LTM) was developed. By interpretion and reconstruction of the gene-reactions association matrix (GRAM) in the original GEM, the relationships among genes associated with the same reaction were included after the transformation. In addition, the stoichiometric matrixes for the metabolism were equalvalents for GEMs before and after LTM. Based on LTM, two novel gene level applications, OptGeneKnock and FastGeneSL, were developed based on two previous methods, OptKnock and FastSL, whose applications were limited to reaction level. These two case studies demonstrated the power of LTM to promote the gene level application of GEMs.Based on the Michaelis-Menten (M-M) equation, genome wide kinetic information of enzymes was integrated as constraints in the latest version of GEM for Saccharomyces cerevisiae. After integration of enzymatic constraints, the GEM was able to precisely predict the Crabtree effect, and could also predict the maximum growth rate of Saccharomyces cerevisiae in mediums using different sugars as carbon source. These proved that the integration of enzymatic constraints would greatly improve the prediction of fluxes with GEMs, and may also help researchers to understand the principle of the microorganism to regulate their protein allocation.Base on the information in databases that are publically available, transcriptional regulatory (TR) and protein-protein interaction (PPI) networks were reconstructed for Saccharomyces cerevisiae. A systematic study was conducted to investigate the relationships between TR and PPI networks and co-expression of genes. Consequently, it was found that when genes shared similar "regulators" in both TR and PPI networks, they were most likey to be co-expressed. The result suggested the TR and PPI networks may be collaborating in modulating the mRNA expression level of target genes. And this finding provided important clue for integration of TR and PPI information in GEMs.In conclusion, this study contributed in applying and expanding GEMs, and exhibited the potential of GEMs in more general applications. In addition, the discovery of the relationship between co-expression and co-regulation facilitated the integration of TR and PPI networks in GEMs. All in all, this study promoted the application of GEMs in fields such as metabolic engineering, bioinformatics and systems medicine. |
PDF Full Text Request