Biosynthesis of plant-derived natural products in the eukaryotic microbe Saccharomyces cerevisiae often faces the issue of the inefficient production due to the poor compatibility between the heterologous genes and chassis cells. In order to improve the biosynthetic efficiency of heterologous production of plant secondary metabolites in S. cerevisiae, people usually do metabolic engineering in and around the heterologous metabolic pathways based on researchers’ experience and mass of trials, which usually consumes a lot of manpower and financial resources. Herein, to further improve the heterologous production of oleanolic acid (OA), a pentacyclic triterpenoid in many plants with several promising pharmacological activities, in a genetically engineered, OA-producing strain S. cerevisiae OA07 effectively, a genome-scale metabolic model of the strain was developed, with the named as Yeast-OA07, and then OptKnock, a flux balance analysis-based pathway design algorithm with bilevel objectives, was utilized to develop in silico gene-knockout strategies to guide the molecular operations in S. cerevisiae OA07. Yeast8-OA07 contained 1133 genes, 2702 metabolites, and 3997 reactions. Five in silico gene-knockout strategies, which were expected to increase OA productivities, were obtained based on the metabolic flux analysis of Yeast8-OA07 through OptKnock. Afterwards, five mutant strains, named as LK1, LK2, LK3, LK4 and LK5, were constructed according to the in silico strategies. It was found that the mutant strain LK2, in which 2-amino-4-hydroxy-6-hydroxymethyl dihydropteridine diphosphokinase-encoding gene FOL1 and formate dehydrogenase-encoding gene FDH1 were deleted, had an OA yield of 125.04 mg·L-1, which was significantlyhigher than the original strain OA07 (89.50 mg·L-1), while the mutant strain LK5, which eliminated paminobenzoic acid synthase-encoding gene ABZ1 and glycine hydroxymethyl transferase-encoding gene SHM1, had an even higher OA yield of 207.37 mg·L-1. Nevertheless, strain LK6, which was developed by integrating the in silico gene-knockout strategies of LK2 and LK5, had a significant decrease of OA production than S. cerevisiae OA07, indicating that in silico knockout strategies do not fit to in vivo iteration directly. Our study provides a novel, efficient method to improve the heterologous production of plant metabolites in microbial cell factories.