CDI-Type I:Engineering Massively Parallelized Fluidic Processors: From Data to Predictive Models to Functional Designs

目前，微流控装置可以产生数百万纳米级的液滴，从而实现高通量的生物分析。考虑到生物分析的多步骤性质，这些液滴通常需要通过流体通道网络穿梭，以便它们可以与其他装载试剂的液滴合并，然后进行分类并最终进行分析。在实验室中实现这一超快流体生物处理器的愿景是一项相当艰巨的任务，因为液滴在相互连接的网络中运动的非线性妨碍了对每个液滴的位置和时间的完全控制。这项工作的主要假设是，计算思维方法可以通过缩小设计空间和产生优化的网络架构解决方案，导致无差错流体处理器精密工程的范式转变。初步数据支持这一假设，激励我们实施计算策略来解决这一设计挑战。首先，将建立处理器基本流体组件的预测模型。其次，预测控制策略将用于解决被动方法与-à-vis主动方法在流体网络中调节液滴轨迹的相对意义。最后，将开发专门的遗传算法（GAs）来优化网络架构，以满足所需的处理器功能。该提案的其他网络方面包括生成和分析大量的数字显微镜数据，捕获网络中液滴的非线性动力学，以及使用最佳可用的计算基础设施从这些丰富的数据中高效地生成知识。因此，拟议的工作涉及多个学科，包括控制理论、系统工程、计算科学、非线性动力学、流体力学、微加工和图像处理。这项研究的结果不仅将推动各种学科的科学和工程前沿，而且还将在生物分析、材料合成、生物传感和疾病诊断等应用领域产生变革性影响。在这项工作中开发的计算工具可以用于分析自然系统中的复杂网络，包括微循环和运输系统。该项目的教育部分包括吸引研究生和本科生进行视觉上引人注目的微流体研究，并在跨学科领域提供最先进的培训-产生一支受过独特培训的劳动力。除了通过会议演示和出版物传播结果外，项目期间生成的数字电影将存储在专用网络服务器上，供其他用户访问，最终建立一个流体处理器架构的数字图书馆。

Currently, microfluidic devices can produce millions of nanoliter-scale droplets allowing high throughput biological analysis. Given the multi-step nature of biological analysis, these droplets often need to be shuttled through a network of fluidic channels so that they can be merged with other reagent-loaded droplets, then sorted and eventually analyzed. Realizing this vision of an ultrafast fluidic bioprocessor in the laboratory is quite a daunting task because non-linearity in the motion of droplets through interconnected networks precludes full control over the position and timing of each and every droplet. The principal hypothesis of this work is that computational thinking approaches can lead to a paradigm shift in precision engineering of error-free fluidic processors by narrowing down the design space and yielding optimized solutions of network architecture. Preliminary data supports this hypothesis, motivating us to implement computational strategies to address this design challenge. First, predictive models of the basic fluidic components of a processor will be built. Second, predictive control strategies will be used to address the relative significance of passive approaches vis-à-vis active methods to regulate droplet trajectories in fluidic networks. Finally, specialized genetic algorithms (GAs) will be developed to optimize network architecture for desired processor functionality. Additional cyber aspects of the proposal include generation and analysis of tremendous amount of digital microscopy data capturing the non-linear dynamics of droplets in networks and efficient knowledge generation from this abundant data using the best available computational infrastructure. Thus, the proposed work cuts across several disciplines including control theory, systems engineering, computational science, non-linear dynamics, fluid mechanics, microfabrication and image processing. The results from this study will not only advance scientific and engineering frontiers in a variety of disciplines but will also lead to transformative impact in applications related to biological analysis, material synthesis, biosensing and disease diagnostics. The computational tools being developed in this work can be adapted to analyze complex networks in natural systems including microcirculation and transportation systems. Educational component of the project includes drawing graduate and undergraduate students to the visually striking microfluidics research and providing state-of-the-art training in interdisciplinary areas - yielding a workforce that is uniquely trained. In addition to disseminating the results through conference presentations and publications, the digital movies generated during the project will be stored on dedicated network servers for access to other users to eventually build a digital library of fluidic processor architectures.