Single-cell feedback, optogenetics, and deep learning to control gene expression in bacteria

细菌群落中的异源基因表达变化被认为产生具有特定生存益处的亚群，例如抗生素耐受性。 目前，科学家们几乎没有工具来以受控的方式模拟和研究这些基因表达事件。 该项目的总体目标是开发基于光的工具，用于精确驱动靶细菌细胞中的基因表达。 这些工具有可能揭示基因表达的异质性变化如何影响细菌细胞亚群的存活。除了这些科学进步，研究计划还辅以教育和推广活动。其中包括与儿童翻滚科学播客合作，为年轻观众（6 - 12岁）制作节目。该项目还为本科生提供培训机会，并包括与K-12科学教育有关的额外推广工作。 该项目的技术目标是开发在单细胞中精确驱动基因表达的方法，并在细菌应激反应的背景下测试这些技术。 为了实现这一目标，研究人员将使用高通量方法，该方法使用自动显微镜，光遗传学和微流体来实现实时反馈控制。 细胞过程是噪声和非线性的，并且可能涉及遗传上相同的细胞之间的延迟和差异响应。 因此，传统的控制算法要么不够精确，要么计算太昂贵，不适合大规模的基因调控网络动态分析。 该提案采用了一种新方法，利用了基于深度学习模型的新兴控制算法。研究人员将进行高通量实验，将光遗传学光信号应用于许多单细胞，以生成用于训练深度学习模型的数据集。 然后，深度学习模型将用于使用深度模型预测控制在单细胞中实施反馈控制，以快速预测要应用的最佳光遗传刺激。 这些工具的最终效用在于它们能够精确控制生物网络中的基因。为了测试这种潜力，研究人员将重点放在酸应激反应网络上。 通过使用反馈控制来对网络调节器施加动态配置文件，研究人员将研究信号如何在网络内传播，以及细胞如何使用基因表达动态来对冲突然出现的压力。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

Heterogenous gene expression changes in bacterial communities are thought to produce subpopulations with particular survival benefits such as antibiotic tolerance. Currently, scientists have few tools to mimic and study these gene expression events in a controlled manner. The overall goal of this project is to develop light-based tools for precisely driving gene expression in targeted bacterial cells. The tools have the potential to reveal how heterogenous changes in gene expression impact the survival of subpopulations of bacterial cells. In addition to these scientific advances, the research program is complemented by educational and outreach activities. These include partnering with the Tumble Science Podcast for Kids to create episodes targeted for young audiences (ages 6-12). The project also provides training opportunities for undergraduate students and includes additional outreach efforts related to K-12 science education. The technical goal of this project is to develop methods to precisely drive gene expression in single cells, and to test these techniques in the context of bacterial stress response. To achieve this, the researchers will use a high-throughput approach that uses automated microscopy, optogenetics, and microfluidics to implement real-time feedback control. Cellular processes are noisy and nonlinear, and can involve delays and differential responses between genetically identical cells. Thus, traditional control algorithms are either not accurate enough or too computationally expensive to be appropriate for large-scale analysis of gene regulatory network dynamics. This proposal takes a new approach, taking advantage of emerging control algorithms based on deep learning models. The researchers will conduct high-throughput experiments where optogenetic light signals are applied to many single cells to generate datasets for training a deep learning model. The deep learning model will then be used to implement feedback control in single cells using Deep Model Predictive Control to rapidly predict the optimal optogenetic stimulus to apply. The ultimate utility of these tools is in their ability to precisely control genes within biological networks. To test this potential, the researchers will focus on the acid stress response network. By using feedback control to impose dynamic profiles on network regulators, the researchers will study how signals propagate within the network and how cells use gene expression dynamics to hedge against the sudden appearance of stress.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.