During the planning and early execution of the Human Genome Project, the genome analysis problems of gene finding and gene functional identification were viewed as very hard problems. Who would have dared to suggest at that time that we might one day be able to computationally generate a quantitative metabolic model of an organism from its sequenced genome? Yet that goal has been achieved, albeit with much room for improving the accuracy of such models. This talk will describe bioinformatics methods for inferring the reactome of an organism from its annotated genome, and for inferring a steady-state quantitative metabolic model from the reactome. Gap filling of models, such as identifying a minimal number of missing metabolic reactions, is a key part of such methods because incomplete metabolic networks do not yield solvable metabolic models. We also consider methods for validating steady-state metabolic models, and applications of metabolic models. One application is the inference of alternative minimal nutrient sets of an organism from its metabolic network, which may provide insights into unculturable organisms.