Optimizing Ethanol Production in Saccharomyces cerevisiae at Ambient and Elevated Temperatures through Machine Learning-Guided Combinatorial Promoter Modifications

Regulatory mechanisms underlying yeast chemical stress response and development of robust strains for bioproduction

Yeast is widely studied in producing biofuels and biochemicals using renewable biomass. Among various yeasts, Saccharomyces cerevisiae has been particularly recognized as an important yeast cell factory. However, economic bioproduction using S. cerevisiae is challenged by harsh environments during fermentation, among which inhibitory chemicals in the culture media or toxic products are common experiences. Understanding the stress-responsive mechanisms is conducive to developing robust yeast strains. Here, we review recent progress in mechanisms underlying yeast stress response, including regulation of cell wall integrity, membrane transport, antioxidative system, and gene transcription. We highlight epigenetic regulation of stress response and summarize manipulation of yeast stress tolerance for improved bioproduction. Prospects in the application of machine learning to improve production efficiency are also discussed.

Engineering Saccharomyces cerevisiae for application in integrated bioprocessing biorefineries

After decades of research and development, no organism - natural or engineered - has been described that can produce commodity products through direct microbial conversion to meet industry demands in terms of rates and yields. Variation in lignocellulosic biomass (LCB) feedstocks, the lack of a widely applicable pretreatment method, and the limited economic value of energy products further complicates second-generation biofuel production. Nevertheless, the emergence of advanced genomic editing tools and a more comprehensive understanding of yeast metabolic systems offer promising avenues for the creation of yeast strains tailored to LCB biorefineries. Here, we discuss recent advances toward developing yeast strains that could convert different LCB fractions into a series of economically viable commodity products in a biorefinery.

Building a machine-learning model to predict optimal mevalonate pathway gene expression levels for efficient production of a carotenoid in yeast

Simultaneous modification of the expression levels of many metabolic enzyme genes results in diverse expression ratios of these genes; however, the relationship between gene expression levels and chemical productivity remains unclear. However, clarification of this relationship is expected to improve the productivity of useful chemicals. Supervised machine learning is considered to be an effective means to clarify this relationship. In this study, to improve the productivity of carotenoids in yeast Saccharomyces cerevisiae, we aimed to build a machine-learning model that can predict the optimal gene expression level for carotenoid production. First, we obtained data on the expression levels of mevalonate pathway enzyme genes and carotenoid production. Then, based on these data, we built a machine-learning model to predict carotenoid productivity based on gene expression levels. The prediction accuracy of 0.6292 (coefficient of determination) was achieved using the test data. The maximum predicted carotenoid productivity was 4.3 times higher in the engineered strain than in the parental strain, suggesting that the expression levels of the mevalonate pathway enzyme genes tHMG1 and ERG8 have a particularly large impact on carotenoid productivity. This study could be one of the important achievements in addressing the uncertainty of genotype-phenotype correlations, which is one of the challenges facing metabolic engineering strategies.