Bridging the gap between theory and experiment in paramagnetic NMR analysis

拟议的研究项目的目标是通过集成高水平量子化学，改进的pNMR位移和弛豫模型以及机器学习来创新顺磁NMR数据分析。提出的基准研究将能够定义针对系统范围量身定制的不同方法的准确性。在这个项目中使用机器学习将使应用于各种分子的pNMR数据分析的自动化向前迈进一步。在现代世界，最大的挑战往往出现在单个分子甚至单个原子的最小尺度上。因此，获取有关该规模系统的信息变得越来越重要。有许多分析技术，如x射线衍射，使我们能够一瞥分子结构，但大多数都需要固体状态的样品。分子结构及其在溶液中的动力学性质的鉴定通常是通过核磁共振（NMR）光谱来完成的。几年前，当主要的核磁共振软件生产商发布了他们的计算机辅助结构解析（CASE）产品时，简单抗磁性有机分子的核磁共振数据分析的进展已经达到了完全自动化的里程碑。现在，只需点击一下按钮，用户就可以分配核磁共振光谱中的所有峰，并获得样品中物种的分子结构。当涉及到顺磁分子或所谓的pNMR时，分析能力的对比是惊人的。科学界对于用什么方程来描述pNMR峰的位置和线宽仍然存在争论。pNMR数据分析目前处于与20世纪初的x射线衍射晶体学类似的阶段，当时一个分子的表征可以作为单独的手稿发表。然而，顺磁性分子形成了一类重要的化合物。特别是顺磁性金属配合物，它被用作蛋白质结构和动力学解释的标签，这些信息构成了现代药物设计的基础。它们也被用作造影剂或磁共振诊断的属性响应探头。

The goal of the proposed research project is to innovate the paramagnetic NMR data analysis by integrating high-level quantum chemistry, improved models of pNMR shift and relaxation and machine learning. The proposed benchmark study would be able to define accuracy of different approaches tailored to range of systems. The employment of machine learning in this project would make a step forward to the automation of pNMR data analysis applied to wide variety of molecules.In the modern world, the biggest challenges often appear in the smallest scale of a single molecule or even a single atom. Hence, obtaining information about systems on that scale becoming more and more important. There are number of analytical techniques, such as X-Ray diffraction, that allow us to get a glimpse of molecular structure but majority of them require samples in the solid state. Identification of molecular structure and its dynamic properties in solution is often done by Nuclear Magnetic Resonance (NMR) spectroscopy. Advances in NMR data analysis for simple diamagnetic organic molecules have reached a milestone of a complete automation several years ago, when major NMR software producers released their Computer-Assisted Structure Elucidation (CASE) products. Now, with a hit of a button a user can assign all the peaks in the NMR spectra and get a molecular structure of species in the sample. When it comes to paramagnetic molecules or so-called pNMR, the contrast in analytic capabilities is striking. There are still debates in the scientific community on what equations describe pNMR peaks positions and linewidths. pNMR data analysis is currently at a similar stage as X-Ray diffraction crystallography in the beginning of XX century, when characterisation of one molecule could be published as a separate manuscript. Yet, paramagnetic molecules form an important class of compounds. In particular, paramagnetic metal complexes, which are employed as tags for protein structure and dynamics elucidation - information that forms the basis for modern drug design. They are also used as contrast agents or property responsive probes for magnetic resonance diagnostics.