Collaborative Research: FET: Small: Hierarchical Computational Framework for large scale RNA Design Pathway Discovery through Data and Experiments

由于在诊断学、治疗学、合成生物学、生物材料和分子编程等不同科学领域的应用，RNA纳米结构设计受到了前所未有的关注。然而，设计和合成具有更高稳定性、可编程几何结构和可控制功能的长RNA分子是一项令人难以置信的具有挑战性的任务。大RNA设计的困难在于它们的长序列和复杂的碱基之间的相互作用。此外，一旦结构设计完成，进行实验既耗时又昂贵。开发一个具有高效设计算法和工具的平台，实现高效、准确的RNA设计是非常有价值的。该项目将促进国家健康繁荣和福利，为设计和合成具有所需功能和改进稳定性的长RNA提供所需的知识。这些RNA将在药物输送和癌症治疗等应用中产生重要的潜在影响。该团队将开发新的计算方法，以更有效和更知情的方式支持下一代纳米结构的发现。它还将提高对表征大规模RNA序列折叠的基本规则的理解和知识。该项目将涉及算法开发和实验活动。因此，教育材料将在科学和工程项目中开发，将不同的学生群体聚集在一起。大多数现有的RNA设计算法专注于保守的、自然进化的3D RNA基序。这些算法使用了“块”的概念，它由~10nts的核苷酸组成，并研究块内和块之间核苷酸对之间的可能结合(块驱动方法)。目前的方法存在预测大RNA分子折叠(200nts)的精度较低的问题。这是一个关键问题，因为迫切需要产生更长的序列来充分利用RNA功能，如催化、基因调控、大型机器中蛋白质的组织以及它们在材料和生物医学科学中的应用。几个挑战使得设计大规模RNA结构的任务变得困难。例如，即使RNA化合物不是最小自由能构型，它们也可以是稳定的。此外，实验表明，对于相同的RNA序列，如何存在不同的构型，并具有不同的相关最小自由能水平。因此，有必要提出与奖励函数无关的实验和计算方法，例如嵌入数据驱动信息以确定RNA构型存在的可能性。这一多学科项目将解决RNA结构设计发展中的两个主要挑战。(I)在不明确知道奖励函数的情况下执行优化。在这个项目中，将提出由经验发展的专家驱动的优化概念。(2)将(I)中的方法推广到高维情况。采用生物启发的瓷砖概念来创建一个计算高效的算法框架来生成和探索瓷砖，这将使用专家驱动的瓷砖链推出进行评估。随后将使用现有的RNA数据库对生成的算法进行验证。将通过算法框架提出和构建新的RNA构建块，作为设计尺寸和复杂性不断增加的单链RNA折纸结构的辅助工具进行验证，这些结构可能会与天然RNA机器或设计的DNA纳米结构相媲美。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。

RNA nanostructure design has received unprecedented attention due to the number of emerging applications in different scientific fields, such as diagnostics, therapeutics, synthetic biology, biological materials, and molecular programming. However, the design and synthesis of long RNA molecules with improved stability, programmable geometries, and controllable functions is an incredibly challenging task. The difficulties of large RNA design are due to their long sequences and complex interactions between bases. In addition, once a structure is designed, conducting experiments is time-consuming and expensive. It is invaluable to develop a platform with effective design algorithms and tools for RNA design with high efficiency and accuracy. This project will advance national health prosperity and welfare, providing the required knowledge for the design and synthesis of long RNAs with the desired functionalities and improved stabilities. These RNAs will have an important potential impact in applications such as drug delivery and cancer therapy. The team will develop new computational methods enabling support to the discovery of next-generation nanostructure in a more efficient and informed manner. It will also improve the understanding and knowledge of fundamental rules that characterize the folding of large-scale RNA sequences. The project will involve algorithm development along with experimental activities. As a result, educational material will be developed across science and engineering programs bringing a diverse group of students together.Most existing RNA-design algorithms focus on conserved, naturally evolved 3D RNA motifs. These algorithms employ the idea of a “block”, which consists of nucleotides (nts) at the scale of ~10nts, and investigate the possible bindings among nucleotides pairs within and between the blocks (block-driven approach). Current approaches suffer from the low accuracy for prediction of large RNA molecules folding (200 nts). This is a critical issue because there is a compelling need to generate longer sequences to fully exploit RNA functionalities such as catalysis, gene regulation, organization of proteins in large machineries, and their use in material and biomedical sciences. Several challenges make the task of designing large-scale RNA structures hard. As an example, RNA compounds can be stable even when they are not minimum free-energy configurations. Also, experiments have shown how alternative configurations can exist for the same RNA sequence, with different associated levels of minimum free energy. It is therefore necessary to come up with approaches, experimental as well as computational, that can be agnostic to a reward function, e.g., embed data-driven information to determine the likelihood of an RNA configuration to exist. This multidisciplinary project will tackle two main challenges for the development of the design of RNA structures. (i) Perform optimization without explicit knowledge of a reward function. The concept of optimization driven by empirically developed experts will be developed in this project. (ii) Scale the methods in (i) to high-dimensional cases. A bio-inspired concept of tile is employed to create a computationally efficient algorithmic framework to generate and explore tiles, which will be evaluated using expert-driven rollout over chains of tiles. The produced algorithms will be subsequently validated using existing RNA databases. New RNA building blocks will be proposed and constructed through the algorithmic framework, to be validated as an assistant tool to the design of single-stranded RNA origami structures with increasing size and complexity that could potentially rival the natural RNA machineries or designer DNA nanostructures.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.