New methods for computational modeling of RNA structures

RNA分子在几乎所有的细胞过程中都在基因表达和 监管。毫不奇怪，新兴的生物医学进步，如精确医学和合成生物学，都 指出RNA是中央调节器和信息载体。最近，人们意识到了使用RNA的潜力 为了干预基因表达，科学家们成功地开发出了Onpattro，这是FDA批准的第一种基于fi的核糖核酸 2018年8月接受治疗。 了解RNA的功能和治疗应用需要有关RNA结构的知识。不幸的是- 目前，已知结构的数量只占需要确定的一小部分。这一差距 必须通过计算方法来关闭。此外，RNA分子是高度带电的聚阴离子，并且 正电荷，如金属离子，与核糖核酸结合，是核糖核酸结构的组成部分。地点和方式 金属离子与核糖核酸的相互作用可以直接影响核糖核酸的结构和功能以及核糖核酸与药物的相互作用。 在NIH的持续支持下，我们已经开发了系统的计算工具，用于 RNA结构、折叠稳定性、动力学和金属离子效应的预测。这些工具带来了丰硕的成果 在病毒学、微生物学、基因治疗、RNA生物技术以及各种基于RNA的治疗性疾病中的应用 有征兆。然而，尽管经过了十多年的努力，计算RNA生物学中的许多关键问题仍然存在： 非Watson-Crick相互作用的从头预测，大RNA的结构预测，有效整合 将低温电子显微镜和核磁共振等实验数据用于结构预测和金属离子效应的模拟。 影响很大。在这笔拨款中，经过15年的初始物理模型的开发，我们建议以上述为目标 和其他紧迫问题，使用一种根本不同的方法，通过系统地开发数据驱动 (如深度学习)或混合数据驱动/基于物理的模拟方法。新的方法是更多的- 随着实验数据量的不断增加和对fi更高效、更可靠的迫切需求 数据解释的计算工具，特别是结构确定实验。我们将使用ex- 周边数据库，如RNA-Puzzles数据库，PDB，EMDataBank，BMRB，用于大型基准 测试，以及由我们久负盛名的合作者收集的生化和核磁共振数据，以进行深入和互动 关于各种实验的信息，如丙型肝炎病毒基因组RNA和艾滋病毒PBS系统。我们的目标，如果成功了- 完全完成，将立即影响结构确定等实验，包括冷冻-EM和 基于核磁共振的结构确定，金属离子中心的识别和fi结构的合理设计 治疗应用。

RNA molecules play fundamental roles in nearly all cellular processes at the level of gene expression and regulation. Not surprisingly, emerging biomedical advances such as precision medicine and synthetic biology all point to RNA as the central regulators and information carriers. Recently, realizing the potential of using RNA to intervene gene expression, scientists successfully developed Onpattro, the first FDA-approved RNA-based therapy in August 2018. Understanding RNA function and therapeutic applications requires knowledge about RNA structure. Unfortu- nately, currently, the number of the known structures is a small fraction of what need to be determined. This gap has to be closed by computational methods. Furthermore, an RNA molecule is a highly charged polyanion and positive charges such as metal ions bind to an RNA and for an integral part of an RNA structure. Where and how metal ions interact with an RNA can directly impact RNA structure and function as well as RNA-drug interactions. Continuously supported by NIH for over 15 years, we have developed systematic computational tools for the predictions of RNA structures, folding stability, kinetics, and metal ion effects. These tools have led to fruitful applications in virology, microbiology, gene therapy, RNA biotechnology, and various RNA-based therapeutic de- signs. However, despite over decade of efforts, many critical issues in computational RNA biology still remain: de novo prediction of non-Watson-Crick interactions, structure prediction for large RNAs, effective incorporation of experimental data such as cryo-EM and NMR data into structure prediction, and modeling of metal ion ef- fects. In this grant, after 15 years of developing an initio physics-based models, we propose to target the above and other pressing issues using a fundamentally different approach by systematically developing data-driven (such as deep-learning) or hybrid data-driven/physics-based simulation methods. The new approaches are mo- tivated by the increasing amount of experimental data and the pressing need to have more efficient and reliable computational tools for data interpretation, especially for structure determination experiments. We will use ex- perimental database, such as RNA-Puzzles database, PDB, EMDataBank, BMRB, for large-scale benchmark tests, and biochemical and NMR data collected by our well-established collaborators for in-depth and interactive information about various experiments such as HCV genomic RNAs and HIV PBS systems. Our goal, if success- fully accomplished, will immediately impact experiments such as structure determination, including cryo-EM and NMR-based structure determination, identification of metal ion sites, and rational design of RNA structures for therapeutic applications.