This dissertation involves several aspects of systems biology. The first part presents the first probabilistic approach to reconstruct genome-scale metabolic networks which integrates sequence homology and context associations. The probabilistic approach not only provides confidence of the annotation of each gene but also suggests likely alternative functions. It is flexible and allows integration of additional types of evidence, such as metabolomic and fluxomic data. The second part describes the reconstruction of the first metabolic network of Plasmodium falciparum suitable for flux balance analysis. The reconstruction is based on available genomic, biochemical, and physiological information, and can be a useful resource to the scientific community studying Plasmodium metabolism. A preliminary validation of the reconstructed metabolic network by comparing gene deletion predictions with published data showed approximately 80% accuracy. The third part is a multi-scale analysis of co-evolution in a metabolic network to understand how the structure and function of a metabolic network influence the pattern of co-evolution of its constituent genes. The fourth part investigates in detail the contribution of gene duplicates to back-up against deleterious human mutations, which can be used in search of disease associated mutations. The last part describes an automatic policing method to detect biochemical misannotations using genomic context correlations.