(281c) Machine-Leaning and Metabolite Biosensor Assisted Construction of Highly Efficient Saccharomyces Cerevisiae Cell Factory

Saccharomyces cerevisiae is an excellent host for heterologous bioproduction of natural products, including therapeutics, chemicals and fuels. The rapidly advanced method to rewire yeast metabolism has allowed more designs to be implemented in genome with high speed and precision. Compared to the methods developed for design and build, the methods to test the performance of newly engineered strains in high-throughput are lagging far behind. To construct highly efficient cell factories, we developed an efficient Machine-learning workflow in conjunction with YeastFab Assembly strategy (MiYA) for combinatorial optimizing the large biosynthetic genotypic space of heterologous metabolic pathways in S. cerevisiae. MiYA utilizes the genotypic and phenotypic information of a small library of representative strains as training dataset to predict possible designs with high efficiency. What’s more, the metabolite biosensors, which could transfer the concentration of target metabolite into detectable signal, were adopted for screening highly efficient cell factory designs in multiplex. Due to the lack of available biological parts, the complexity of the transcriptional regulation network, and the lack of clear understanding of the transcriptional regulation network in yeast, we also developed a standard workflow to finely tune the transfer function of metabolite biosensor to assist the transplantation of prokaryotic transcription factors into yeast. The combination of machine learning and metabolite biosensor would accelerate the design-build-test-learn cycle.