FET: SMALL: New Abstraction and Design Automation for Complex Computations with DNA Using Fractional Coding

在复杂的生物分子电路中使用基于DNA的技术是合成生物学中一个快速发展的领域，有可能扰乱人们设计智能药物或提供靶向治疗的方式。鉴于它们的生物学特性和体内兼容性，基于合成DNA的计算电路的其他应用现在无法预见，肯定会接踵而至。生物技术的最新进展显示了实现基于DNA的复杂电路的潜力。然而，随着这些电路的复杂性增加，它们变得更慢，更难有效地设计。正如硅基技术所显示的那样，新的设计方法和自动化工具对于该领域的进步至关重要。该项目旨在为使用分数编码的快速和复杂的分子计算电路开发一个系统的设计框架和新的方法，并研究使用该框架来高效和可扩展地实现人工神经网络(ANN)。随着设计自动化框架的开发，将产生一个称为FUNDNA的配套软件，它将数学函数映射到DNA反应。这种免费提供的开源软件及其教程和文档将改变生物设计自动化研究，使分子电路能够高效地设计用于复杂计算的数学表示，而不需要生物/化学技能。该工具将使创建新型复杂DNA计算电路的速度更快、成本更低、更容易获得。该项目包括两项主要的教育举措，目的是扩大对科学、技术和经济教育的参与。首先，工程领域代表性不足的学生将通过肯塔基州-西弗吉尼亚州LSAMP计划、全国黑人工程师协会肯塔基大学(英国)学生分会和英国学生支持服务机构进行培训，以便参与研究项目。将开发一门新课程，分子编程和DNA计算电路，以整合这项研究的知识和结果，并帮助培训学生在DNA计算领域。其次，将成立一个iGEM(国际基因工程机器)分会，这是肯塔基州的第一个iGEM团队，以共享资源并在国际合成生物学比赛中竞争。该项目将使高效分子电路的系统化设计成为可能，以便快速计算复杂的数学函数。分子计算电路的设计将是系统化的，因为该项目将开发一个新的设计抽象级别，接受所需计算的数学表示。它将创建一个端到端的编译器，将数学表示传输到化学反应网络，然后生成候选DNA反应和序列。所开发的电路将是快速的，因为它们基于分数编码工作。对于传统的分子编码，每个输入/输出由分子浓度表示，而对于分数编码，输入/输出是分子浓度对的比率。基于传统编码的分子电路需要等待分子的最终(平衡)浓度，而对于基于分数编码的电路，比率到达最终值的速度比分子浓度快得多。该项目将通过两种方法将分子计算电路的计算能力扩展到复杂计算：1)将所有电子随机计算逻辑电路映射到分子电路；2)通过分子反应实现有限状态马尔可夫链。最后，该项目将开发利用分子反应进行神经网络的新方法。我们的目标是通过三个目标展示分数编码如何改进分子计算，特别是DNA计算。1)开发一个设计自动化框架和一个称为FUNDNA的配套软件，该软件将数学函数映射到DNA反应。2)研究基于分数编码的神经网络等复杂计算。3)使用分数编码的基本分子计算电路的实验DNA实现。该项目将从高层次的数学或算法表示出发，引入并尝试实现分子电路的自动化设计思想，将对生物设计自动化领域产生重大影响。虽然在不同类型的数学函数的分子计算方面已经有了先前的工作，但还没有提出在分子电路中计算复杂的一般数学函数的系统方法。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The use of DNA-based technologies in complex biomolecular circuits is a quickly developing field of synthetic biology that has the potential to disrupt how one designs smart drugs or delivers targeted treatments. Given their biological nature and in-vivo compatibility, other applications for synthesized DNA-based computational circuits that cannot be foreseen now will surely follow. Recent advances in biotechnology have shown the potential for implementing complex DNA-based circuits. However, as the complexity of these circuits increases, they become slower and more difficult to be designed efficiently. As shown in silicon-based technology, new design approaches and automation tools are essential for progress in the field. This project aims to develop a systematic design framework and new approaches for fast and complex molecular computing circuits using fractional coding and to investigate the use of this framework for the efficient and scalable implementation of artificial neural networks (ANNs). Along with the development of the design automation framework, a companion software, called FUNDNA, that maps mathematical FUNctions to DNA reactions will be produced. This freely available, open-source software with its tutorials and documentation will transform bio-design automation research by enabling efficient design of molecular circuits for complex computations with mathematical representation without requiring biological/chemical skills. The tool will make the creation of novel complex DNA computing circuits faster, cheaper, and more accessible. This project includes two major educational initiatives with the goal of broadening participation in STEM. First, students who are underrepresented in engineering fields will be recruited through the Kentucky-West Virginia LSAMP program, the University of Kentucky (UK) Student Chapter of the National Society of Black Engineers, and UK Student Support Services to be trained to work on the research project. A new course, Molecular Programming and DNA Computing Circuits, will be developed to integrate the knowledge and findings of this research and to help train students in the area of DNA computing. Second, an iGEM (International Genetically Engineered Machine) chapter, the first iGEM team in the state of Kentucky, will be established to share resources and compete in international synthetic biology competitions. This project will enable systematic design of efficient molecular circuits for fast computation of complex mathematical functions. The design of molecular computing circuits will be systematic because the project will develop a new level of design abstraction that accepts mathematical representation of desired computations. It will create an end-to-end compiler that transfers the mathematical representation to chemical reaction network, and then generates candidate DNA reactions and sequences. The developed circuits will be fast because they work based on fractional coding. While for traditional molecular encoding each input/output is represented by a molecular concentration, for fractional coding inputs/outputs are ratios of pairs of molecular concentrations. Molecular circuits based on traditional encoding need to wait for the final (at equilibrium) concentrations of molecules, whereas for circuits based on fractional coding, the ratios reach their final values much faster than molecular concentrations. This project will expand the computational power of molecular computing circuits to complex computations through two approaches: 1) it will map all electronic stochastic computing logic circuits to molecular circuits, 2) it will implement finite state Markov chains by molecular reactions. Finally, the project will develop new approaches for performing ANNs with molecular reactions. The goal is to demonstrate how fractional coding can improve molecular computation, and DNA computing in particular, through three objectives. 1) Development of a design automation framework and a companion software, called FUNDNA, that maps mathematical FUNctions to DNA reactions. 2) Investigation of complex computations such as neural networks based on fractional coding. 3) Experimental DNA implementation of basic molecular computing circuits using fractional coding. This project will have a significant impact on the field of bio-design automation because it will introduce and attempt to realize the idea of automated design of molecular circuits, starting from high level mathematical or algorithmic representation. While there is prior work in molecular computation of different sorts of mathematical functions, no systematic method has been proposed for computing complex general mathematical functions within a molecular circuit.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.