Plant natural products (PNPs) play important roles across diverse fields of human society including pharmaceutical and agriculture. Compared to traditional plant extraction or chemical synthesis, metabolic engineering in microbial systems such as yeast provides a more rapid, cost-effective, efficient, and environmental-friendly approach to produce these valuable PNPs. The first part of this thesis demonstrates the complete biosynthesis of berberine, a typical benzylisoquinoline alkaloid with multiple bioactivities, in yeast with the highest titer reported (1.08 mg/L). The engineered yeast strain also showed high efficiency to incorporate halogenated tyrosine for the synthesis of unnatural BIA derivatives that have higher therapeutical potentials. In addition to strain engineering, the second part of the thesis presents an orthogonal strategy to engineer metabolic flux via artificial cell wall with selective permeability, which further enhanced the production of PNPs, for example, berberine by 2.58-fold. Yeast can serve as a versatile platform for not only biomanufacturing PNPs, but also studying their biosynthetic mechanisms and discovering unknown PNP biosynthetic pathways. The third part of this thesis reveals the regulatory mechanisms of isoflavone biosynthetic pathway. By reconstructing multiple complex variants with a deactivated soybean isoflavone synthase (GmIFS2) in yeast, we found that GmIFS2-mediated protein-protein interactions (PPIs) can augment and redirect metabolic flux between two competing pathway branches producing deoxy(iso)flavonoids and (iso)flavonoids, and further confirmed the mechanism by identifying two novel isoflavone O-methyltransferases (GmIOMTs) interacting with GmIFS2. The deep involvement of PPI in PNP biosynthetic pathways inspired the final part of this thesis, where we established a yeast-based high-throughput platform for detecting PPI of plant enzymes to predict undiscovered pathways. As a proof of concept, we screened a library of 331 medium-chain dehydrogenases/reductase (MDRs) involved in the less studied pathway of monoterpene indole alkaloids (MIAs). 47 of them were screened out and two novel enzymes, which share low sequence similarity with known enzymes, were biochemically identified from them, demonstrating a novel strategy to predict PNP pathways. Overall, this dissertation highlights the huge engineering value of yeast platform to produce PNPs and provides insights on pathway discovery with synthetic biology tools in yeast.