For many fungal genomes of biotechnological interest, the combination of large-scale sequencing projects and in-depth experimental studies has made it feasible to undertake metabolic network reconstruction. An excellent representative of this new class of organisms is Yarrowia lipolytica, an oleaginous yeast studied experimentally for its role as a food contaminant and its use in bioremediation and cell factory applications. As one of the hemiascomycetous yeasts completely sequenced in the Génolevures program it enjoys a high quality manual annotation by a network of experts. We have developed a method of semi-automatic reconstruction of metabolic models, based on the prediction of conservation of enzymatic activities between two species. Following our reconstruction protocol, we extrapolated a Y.lipolytica genome-scale metabolic model (called iNL705) from an existing S. cerevisiae model (iIN800). This draft model was curated by a group of experts in Y. lipolytica metabolism, and iteratively improved and validated through comparison with experimental data by flux balance analysis. In order to design better automatic methods, we formalized the steps of expert manual curation as an algebra of edit operations on metabolic models. The cycle of iterative refinements is represented as transformations of the automatically produced draft model, with an evaluation of improvement in accuracy after every step. This study underscores the particular challenges of metabolic model reconstruction for eukaryotes. The experience acquired with our methods and formalizations should prove useful for similar reconstruction efforts.