Next-Generation NMR Crystallography Through Ab Initio Structure Refinement

加州大学河滨分校的Gregory Beran获得了化学部化学理论、模型和计算方法计划颁发的奖项，以开发新的计算工具，促进通过核磁共振(核磁共振)光谱确定三维晶体结构。分子晶体结构知识在制药学和许多其他化学领域中是必不可少的。同一分子的不同晶体堆积基序(“多晶型”)可能表现出非常不同的性质，而不良多晶型的出现已给患者和制药商带来了重大药物召回和其他严重问题。核磁共振波谱越来越多地被用于确定晶体结构和表征不同的晶型。将核磁共振谱中存在的一组峰转换为三维晶体结构通常涉及对潜在结构进行反复计算建模，以识别其预测的核磁共振谱与实验观察的结构最匹配的结构。该项目开发了计算工具，以绕过这种反复试验的过程，并实现直接的晶体结构精炼，可以自动解决产生实验观察到的核磁共振谱的晶体结构。这些新工具将显著提高核磁共振在解决晶体结构问题上的效用。实现这些目标需要在三个领域取得研究进展。首先，正在开发新的计算实用的电子结构方法来模拟支配晶体堆积的非共价相互作用，以便能够识别良好的初始晶体结构。其次，由于在比较预测光谱和观测光谱时，准确预测给定结构的核磁共振化学位移对于区分正确和不正确的结构指定至关重要，因此正在开发比广泛使用的现有技术更可靠地预测核磁共振化学位移的模型。第三，为了实现直接的核磁共振驱动的晶体结构精细化，正在开发再现分子构象和晶体堆积的变化如何影响化学位移的机器学习模型。成功将带来更大的能力来设计具有所需特性的分子晶体。此外，该奖项还支持当地一所小学的外联活动和制作在线工具，以教育新用户有关计算化学方法的知识。

Gregory Beran of the University of California Riverside is supported by an award from the Chemical Theory, Models and Computational Methods program in the Chemistry Division to develop new computational tools that will facilitate the determination of three-dimensional crystal structures via nuclear magnetic resonance (NMR) spectroscopy. Knowledge of molecular crystal structures is essential in pharmaceuticals and many other areas of chemistry. Different crystal packing motifs ("polymorphs") of the same molecule can exhibit vastly different properties, and occurrences of undesirable polymorphs have caused major drug recalls and other serious problems for patients and pharmaceutical manufacturers. NMR spectroscopy is increasingly used to determine crystal structures and characterize different polymorphs. Translating the set of peaks present in an NMR spectrum into a three-dimensional crystal structure frequently involves trial-and-error computational modeling of potential structures to identify the structure whose predicted NMR spectrum best matches the experimentally observed one. This project develops computational tools to circumvent this trial-and-error process and enable direct crystal structure refinement which can automatically solve for the crystal structure that produces the experimentally observed NMR spectrum. These new tools will significantly improve the utility of NMR for solving crystal structures. Achieving these goals requires research advances in three areas. First, new computationally practical electronic structure methods for modeling the non-covalent interactions that govern crystal packing are being developed to enable identification of good initial crystal structures. Second, because accurate predictions of the NMR chemical shifts for a given structure are essential to discriminating between correct and incorrect structural assignments when comparing predicted and observed spectra, models which predict NMR chemical shifts more reliably than widely used existing techniques are being developed. Third, to enable direct NMR-driven crystal structure refinement, machine learning models that reproduce how changes in molecular conformation and crystal packing impact the chemical shifts are being developed. Success will lead to greater ability to engineer molecular crystals with desired properties. In addition, this award is supporting outreach activities at a local elementary school and production of online tools to educate new users about computational chemistry approaches.