The invention provides a method for information mining of a genome metabolic network preliminary model. The method comprises the steps as follows: as for the KEGG website, extracting a corresponding relation among gene, protein and reaction, that is, determining the GPR relation by adopting a website script semantic analysis technology on the basis of webpage key content priori position information, and establishing an excel table to show the GPR relation information, so as to obtain the genome metabolic network preliminary model. The model constructed by adopting the method unifies the reaction form, and is convenient in comparison with other models and convenient in reference. The method has been applied to construction of a scheffersomyces stipitis CBS6054 genome-scale metabolic network, and compared with the construction of a traditional network preliminary model based on the KEGG database, a great amount of labor and time are saved, and the construction efficiency is greatly improved.