Unified Computation Tools for Natural Products Research

总结 这一拟议更新应用程序的总体目标是进一步推进正在开发的工具， 有效地将几种类型的分析数据与生物测定数据和基因组信息相结合。这将创建一个 一套强大的工具，用于更快，更准确地识别新分子，已知分子的去复制， 从光谱信息直接推断生物活动。在目前的支持期间，我们已经 在开发用于自动注释和鉴定有机分子的非常有用的工具方面取得了实质性进展， 专注于天然产品。全球天然产物社会（GNPS）分子网络分析和 知识传播生态系统在全球近160个国家处理了近160，000个工作岗位，有4- 6，000个 每月提交新的职位，每月访问超过200 000次（大多数访问是为了参考图书馆 访问，检查公共数据和以前的工作，社区在论文中共享超链接），并已成为一个 注释来自不同来源的有机分子的主流工具，特别是在代谢组学中 工作流程。小分子精确识别技术（SMART）的公共网站，一种深度学习模型 用于提供基于1H-13 C HSQC NMR数据的候选结构，于2019年12月上线， 在50个国家提供3000个工作岗位。本提案中开发的所有工具都将成为该分析生态系统的一部分。四 为这项拟议的研究活动做出贡献的实验室已经创建了一个开放和综合的团队， 创造性地创新新的信息学工具，以增强小分子结构注释及其化学性质的推断 和生物特性。我们有四个具体目标：1）完成一套新的 和创新的天然产品分析工具，并将其作为免费资源部署到世界各地， 社区2)通过利用储存库规模的质量来完善分子的结构表征 光谱信息沿着NMR数据和基因组输入。3)要创建一个新的基于SMART的工具， 整合质谱和HSQC NMR数据作为新的深度学习系统的输入，目标是 从而实现更准确的结构预测。4)使用深度学习来增强SMART与生物活性数据 从而使SMART能够根据光谱特征预测分子的活性。数据还将增加 GNPS数据库中的生物测定结合数据。这些目标的另一个结果将是进一步的 天然产物分析数据的数字化，以便它们可以用于本文计划的计算工具， 以及其他工具。这四个具体目标的完成将为精确的 识别新的天然产物结构，并能够推断其与其他类别的结构相关性。 有机分子及其生物学特性。因此，这些新的信息工具将有潜力大大提高 增强小分子药物发现过程。

Summary The overarching goal for this proposed renewal application will be to further advance tools that are in development and to effectively integrate several types of analytical data with biological assay data and genomic information. This will create a powerful set of tools for faster and even more accurate identification of new molecules, dereplication of known ones, and to directly infer biological activities from spectroscopic information. In the current period of support, we have made substantial progress in developing highly useful tools for automatic annotations and identifications of organic molecules, specifically focused on natural products. The Global Natural Products Social (GNPS) Molecular Networking analysis and knowledge dissemination ecosystem has processed almost 160,000 jobs in nearly 160 countries worldwide, has 4-6,000 new job submissions per month and is accessed over 200,000 times a month (majority accessions are for reference library access, inspection of public data and previous jobs that the community shares as hyperlinks in papers), and has become a mainstream tool for the annotation of organic molecules deriving from diverse sources, especially in metabolomics workflows. The public website for Small Molecule Accurate Recognition Technology (SMART), a deep learning model for providing candidate structures based on 1H-13C HSQC NMR data, went live in December 2019 and already has over 3000 jobs in 50 countries. All tools developed in this proposal will become part of this analysis ecosystem. The four laboratories contributing to this proposed research activity have created an open and integrated team that is continuing to creatively innovate new informatic tools to enhance small molecule structure annotations and inference of their chemical and biological properties. We have four specific aims: 1) To complete the development and evaluation of a set of new and innovative tools for natural products analysis, and deploy these as freely available resources for the worldwide community. 2) To refine the structural characterization of molecules through leveraging repository scale mass spectral information along with NMR data and genomic inputs. 3) To create a new SMART-based tool that integrates mass spectrometry and HSQC NMR data as the input for a new deep learning system with the goal of achieving more accurate predictions of structure. 4) To use deep learning to enhance SMART with bioactivity data so as to enable SMART to predict activities of molecules based on spectroscopic features. The data will also augment the GNPS database with biological assay binding data. An additional consequence of these goals will be the further digitization of natural products analytical data so that they can be used in the computational tools planned herein, as well as other tools in the future. Completion of these four specific aims will create new integrated tools for the precise identification of new natural product structures, and enable inference of their structural relatedness to other classes of organic molecules and their biological properties. Thus, these new informatic tools will have the potential to greatly enhance the small molecule drug discovery process.