Collaborative Research: Data-driven engineering of the thermotolerant yeast Kluyveromyces marxianus

新陈代谢包括一系列复杂的反应和控制机制。这使得微生物的行为难以理解和设计。机器学习（ML）识别复杂数据集中的模式和关系。酵母将受到各种努力，以增加其生物化学和生物燃料的产量。机器学习将有助于识别最大化生产的合成生物学方法。此外，教育和培训资源将扩大到K-6课后项目、K-12夏令营和高中生数据科学学院的学生。总体目标是确定对驱动高碳通量到所需的中心代谢物和产物至关重要的基因。深度学习方法（DeepGuide）将用于设计优化的sgRNA文库，以产生遗传多样性。该周期的构建阶段将利用高效的CRISPR-Cas9技术进行基因破坏和调控。生物传感器驱动的检测方法将能够高通量分析宿主遗传对丙二酰辅酶A（聚酮的关键前体）产生的影响。阿贡国家实验室（Argonne National Laboratory）的Agile bioffoundry （ABF）液滴微流体筛选和分析能力将提供丰富的额外信息，将代谢物的产生与宿主遗传学联系起来。测试阶段生成的大型数据集将作为深度学习的输入；一种将基因型与表型联系起来的新算法——深度代谢——将用于预测一组最小的遗传扰动，从而最大化丙二酰辅酶a的生物合成。这种数据驱动的方法将推进微生物工程的深度学习和高通量方法，并可应用于其他战略代谢物。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Metabolism involves a complex set of reactions and control mechanisms. This makes microbial behavior difficult to understand and engineer. Machine learning (ML) identifies patterns and relationships in complex sets of data. A yeast will be subject to varied efforts to increase its yield of biochemicals and biofuels. Machine learning will help identify synthetic biology approaches that maximize production. Also, educational and training resources will be expanded for students in K-6 afterschool programs, in K-12 summer camps, and in a Data Science Academy for high school students. The overall goal is to identify genes critical to driving high carbon flux to a desired central metabolite and product. A deep learning approach – DeepGuide – will be used to design optimized sgRNA libraries to generate genetic diversity. The build stage of the cycle will leverage efficient CRISPR-Cas9 technologies for gene disruption and regulation. A biosensor-driven approach to testing will enable high throughput analysis of the effect of host genetics on the production of malonyl-CoA, a key precursor to polyketides. The droplet microfluidics screening and analysis capabilities of the Agile BioFoundry (ABF) at Argonne National Laboratory will provide a wealth of additional information to link metabolite production with host genetics. The large datasets generated in the testing stage will be used as input for deep learning; a new algorithm linking genotypes to phenotypes – DeepMetabolism – will be used to predict a minimal set of genetic perturbations that maximize malonyl-CoA biosynthesis. This data-driven approach will advance both deep learning and high throughput approaches for microbial engineering and can be applied to other strategic metabolites.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.