EAGER: IIBR Informatics: Deep learning tools for the identification of RNA modifications from direct RNA sequencing data

mRNA修饰作为在多个水平上影响基因表达的调控过程的作用还没有得到很好的研究或理解。目前用于确定RNA修饰的实验工具是费力的、嘈杂的，并且通常不能提供修饰的碱基的精确位置。使用Oxford Nanopore技术的测序提供了超过Illumina测序的多个优势，包括长读段和直接测序RNA而无需扩增的能力，从而降低了覆盖范围的偏倚和发现修饰碱基的潜在能力。由于缺乏用于这项任务的工具，发现修饰碱基的潜力仍然没有得到满足。该项目旨在使全球范围内识别RNA修饰变得更加容易，并帮助揭示150多种不同类型的RNA修饰的生物学作用。拟议研究的挑战在于已知修改的碱基数量相对较少，需要巧妙设计足够大的标记数据集，并且需要使用深度学习训练算法，尽管数据集相对较小，但仍然可以成功。该项目借鉴了深度学习的最新发展，针对几乎没有标记训练样本的任务，开发了新的方法，将用于碱基调用的深度学习架构应用于修饰RNA碱基的调用。拟议的工作将对RNA修饰的研究产生变革性影响，并将使纳米孔测序成为实现这一目的的一站式服务。此外，它有可能导致用于检测RNA结合蛋白的靶标的改进方法，因为用于该任务的几种新方法是基于检测修饰的RNA碱基。Oxford Nanopore不发布其生产基地的代码，将软件称为开源，限制了研究社区扩展其方法以处理修改的能力。作为这项工作的一部分，设计的工具将提供一个灵活的开源替代方案，使纳米孔数据的基础调用取得进展。该奖项反映了NSF的法定使命，并被认为值得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来支持。

The role of mRNA modifications as a regulatory process that affects gene expression at multiple levels is not well studied or understood. Current experimental tools for determining RNA modifications are laborious, noisy, and often do not provide exact locations of modified bases. Sequencing using Oxford Nanopore technology offers multiple advantages over Illumina sequencing including long reads and the ability to directly sequence RNA without the need for amplification, leading to reduced bias in coverage and the potential ability to uncover modified bases. The potential for discovering modified bases is still unfulfilled due to the lack of tools for this task. This project seeks to to make it significantly easier to identify RNA modifications globally and to help uncover the biological roles of the over 150 different types of RNA modifications. The challenge in the proposed research is that of the relatively small number of known modified bases, requiring clever design of sufficiently large labeled datasets, and necessitating the use of deep learning training algorithms that can succeed despite the relatively smaller datasets. The project draws upon recent developments in deep learning for tasks with few available labeled training examples to develop novel ways in which deep learning architectures for base calling can be applied to calling of modified RNA bases.The proposed work will be transformative for research into RNA modifications and will enable the use of nanopore sequencing as a one-stop-shop for this purpose. Furthermore, it has the potential of leading to improved methods for the detection of targets of RNA-binding proteins, as several novel methods for this task are based on detecting modified RNA bases. Oxford Nanopore does not release the code for their production base calling software as open-source, limiting the ability of the research community to extend their methods to handle modifications. The tools designed as part of this work will provide a flexible open-source alternative, enabling progress on base calling of nanopore data.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.