Carl Fleischer, Sean Thomas Holmes, Kirill Levin, Stas L Veinberg, Rob Schurko
Quadrupolar NMR crystallography guided crystal structure prediction (QNMRX-CSP) is a nascent protocol for predicting, solving, and refining crystal structures. QNMRX-CSP employs a combination of solid-state NMR data from quadrupolar nuclides (i.e., nuclear spin > 1/2), static lattice energies and electric field gradient (EFG) tensors from dispersion-corrected density functional theory (DFT-D2*) calculations, and powder X-ray diffraction (PXRD) data; however, it has so far been applied only to organic HCl salts with small and rigid organic components, using 35Cl EFG tensor data for both structural refinement and validation. Herein, the QNMRX-CSP protocol is extended to ephedrine HCl (Eph) and pseudoephedrine HCl (Pse), which are diastereomeric compounds that feature distinct space groups and organic components that are larger and more flexible. A series of benchmarking calculations are used to generate structural models that can be validated against experimental data, and to explore the impacts of the (i) starting structural models (i.e., geometry-optimized fragments based on either a known crystal structure or an isolated gas-phase molecule) and (ii) selection of unit cell parameters and space groups. Finally, we use QNMRX-CSP to predict the structure of Pse in the dosage form Sudafed using only 35Cl SSNMR data as experimental input. This proof-of-concept work suggests the possibility of employing QNMRX-CSP protocols to solve the structures of organic HCl salts in dosage forms – something which is often beyond the capabilities of conventional, diffraction-based characterization methods.
四极核磁共振晶体学指导下的晶体结构预测（QNMRX-CSP）是一种用于预测、解决和完善晶体结构的新兴方案。QNMRX-CSP 将四极核素（即核自旋为 1/2）的固态核磁共振数据、弥散校正密度泛函理论（DFT-D2*）计算得出的静态晶格能和电场梯度（EFG）张量以及粉末 X 射线衍射（PXRD）数据结合起来使用；不过，迄今为止，它只应用于有机成分较少且较硬的有机 HCl 盐，使用 35Cl EFG 张量数据进行结构完善和验证。在本文中，QNMRX-CSP 方案扩展到了盐酸麻黄碱（Eph）和盐酸伪麻黄碱（Pse），这两种非对映化合物具有不同的空间基团和较大且更灵活的有机成分。我们利用一系列基准计算来生成可与实验数据进行验证的结构模型，并探索 (i) 初始结构模型（即基于已知晶体结构或孤立气相分子的几何优化片段）和 (ii) 单胞参数和空间群选择的影响。最后，我们仅使用 35Cl SSNMR 数据作为实验输入，利用 QNMRX-CSP 预测了剂型 Sudafed 中 Pse 的结构。这项概念验证工作表明，可以采用 QNMRX-CSP 协议来解决剂型中有机盐酸盐的结构问题，而这往往是传统的基于衍射的表征方法所无法解决的。
{"title":"Characterization of Ephedrine HCl and Pseudoephedrine HCl Using Quadrupolar NMR Crystallography Guided Crystal Structure Prediction","authors":"Carl Fleischer, Sean Thomas Holmes, Kirill Levin, Stas L Veinberg, Rob Schurko","doi":"10.1039/d4fd00089g","DOIUrl":"https://doi.org/10.1039/d4fd00089g","url":null,"abstract":"Quadrupolar NMR crystallography guided crystal structure prediction (QNMRX-CSP) is a nascent protocol for predicting, solving, and refining crystal structures. QNMRX-CSP employs a combination of solid-state NMR data from quadrupolar nuclides (<em>i.e.</em>, nuclear spin > 1/2), static lattice energies and electric field gradient (EFG) tensors from dispersion-corrected density functional theory (DFT-D2*) calculations, and powder X-ray diffraction (PXRD) data; however, it has so far been applied only to organic HCl salts with small and rigid organic components, using <small><sup>35</sup></small>Cl EFG tensor data for both structural refinement and validation. Herein, the QNMRX-CSP protocol is extended to ephedrine HCl (Eph) and pseudoephedrine HCl (Pse), which are diastereomeric compounds that feature distinct space groups and organic components that are larger and more flexible. A series of benchmarking calculations are used to generate structural models that can be validated against experimental data, and to explore the impacts of the (i) starting structural models (<em>i.e.</em>, geometry-optimized fragments based on either a known crystal structure or an isolated gas-phase molecule) and (ii) selection of unit cell parameters and space groups. Finally, we use QNMRX-CSP to predict the structure of Pse in the dosage form Sudafed using only <small><sup>35</sup></small>Cl SSNMR data as experimental input. This proof-of-concept work suggests the possibility of employing QNMRX-CSP protocols to solve the structures of organic HCl salts in dosage forms – something which is often beyond the capabilities of conventional, diffraction-based characterization methods.","PeriodicalId":76,"journal":{"name":"Faraday Discussions","volume":"19 1","pages":""},"PeriodicalIF":3.4,"publicationDate":"2024-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141501893","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jamie Liam Guest, Esther A. E. Bourne, Martin A. Screen, Mark Richard Wilson, Tran N. Pham, Paul Hodgkinson
Molecular dynamics (MD) simulations and chemical shifts from machine learning are used to predict 15N, 13C and 1H chemical shifts for the amorphous form of the drug irbesartan. The molecules are observed to be highly dynamic well below the glass transition, and averaging over this dynamics is essential to understanding the observed NMR shifts. Predicted linewidths are consistently about 2 ppm narrower than observed experimentally, which is hypothesised to result from susceptibility effects. Previously observed differences in the 13C shifts associated with the two tetrazole tautomers can be rationalised in terms of differing conformational dynamics associated with the presence of intramolecular interaction in one tautomer. 1H shifts associated with hydrogen bonding can also be rationalised in terms of differing average frequencies of transient hydrogen bonding interactions.
{"title":"The essential synergy of MD simulation and NMR in understanding amorphous drug forms","authors":"Jamie Liam Guest, Esther A. E. Bourne, Martin A. Screen, Mark Richard Wilson, Tran N. Pham, Paul Hodgkinson","doi":"10.1039/d4fd00097h","DOIUrl":"https://doi.org/10.1039/d4fd00097h","url":null,"abstract":"Molecular dynamics (MD) simulations and chemical shifts from machine learning are used to predict <small><sup>15</sup></small>N, <small><sup>13</sup></small>C and <small><sup>1</sup></small>H chemical shifts for the amorphous form of the drug irbesartan. The molecules are observed to be highly dynamic well below the glass transition, and averaging over this dynamics is essential to understanding the observed NMR shifts. Predicted linewidths are consistently about 2 ppm narrower than observed experimentally, which is hypothesised to result from susceptibility effects. Previously observed differences in the <small><sup>13</sup></small>C shifts associated with the two tetrazole tautomers can be rationalised in terms of differing conformational dynamics associated with the presence of intramolecular interaction in one tautomer. <small><sup>1</sup></small>H shifts associated with hydrogen bonding can also be rationalised in terms of differing average frequencies of transient hydrogen bonding interactions.","PeriodicalId":76,"journal":{"name":"Faraday Discussions","volume":"26 1","pages":""},"PeriodicalIF":3.4,"publicationDate":"2024-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141501895","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Metal centers are essential for enzyme catalysis, stabilizing the active site, facilitating electron transfer, and maintaining the structure through coordination with amino acids. In this study, K238H-AeL nanopores with histidine sites were designed for the first time as single-molecule reactors for the measurement of single-molecule coordination reactions. The coordination mechanism of Au(Ⅲ) with histidine and glutamate in nano-confined biological nanopores was explored. Specifically, Au(Ⅲ) interacts with the nitrogen (N) atom in the histidine imidazole ring of the K238C-AeL nanopore and the oxygen (O) atom in glutamate to form a stable K238H-Au-Cl2 complex. The formation mechanism of this complex was further validated through single-molecule nanopore analysis, mass spectrometry, and molecular dynamics simulations. By introducing histidine and glutamic acid into different positions within the nanopore revealed that the formation of the histidine-Au coordination bond in the confined space requires a distance within 2.5 Å between the ligand and the central metal atom. By analyzing the association and dissociation rates of single Au(Ⅲ) ions under the applied voltages, it was found that a confined nanopore increased the bonding rate of Au(Ⅲ)-Histidine coordination reactions by around 105 times compared to the bulk solution, and the optimal voltage for single-molecule coordination., providing valuable insights for designing reaction pathways in electrochemical catalysis. This research revealed a novel mechanism for metal coordination and amino acid residues in protein nanoconfined space, highlighting the dynamic interactions between metal ions and amino acid residues and the importance of the confined effect, providing insights for developing efficient, eco-friendly electrocatalytic nanomaterials.
{"title":"Electrochemical Kinetic Fingerprinting of Single-Molecule Cooridations in the Confined Nanopores","authors":"Chaonan Yang, Wei Liu, Haotian Liu, Jichang Zhang, Yi-Tao Long, Yi-Lun Ying","doi":"10.1039/d4fd00133h","DOIUrl":"https://doi.org/10.1039/d4fd00133h","url":null,"abstract":"Metal centers are essential for enzyme catalysis, stabilizing the active site, facilitating electron transfer, and maintaining the structure through coordination with amino acids. In this study, K238H-AeL nanopores with histidine sites were designed for the first time as single-molecule reactors for the measurement of single-molecule coordination reactions. The coordination mechanism of Au(Ⅲ) with histidine and glutamate in nano-confined biological nanopores was explored. Specifically, Au(Ⅲ) interacts with the nitrogen (N) atom in the histidine imidazole ring of the K238C-AeL nanopore and the oxygen (O) atom in glutamate to form a stable K238H-Au-Cl2 complex. The formation mechanism of this complex was further validated through single-molecule nanopore analysis, mass spectrometry, and molecular dynamics simulations. By introducing histidine and glutamic acid into different positions within the nanopore revealed that the formation of the histidine-Au coordination bond in the confined space requires a distance within 2.5 Å between the ligand and the central metal atom. By analyzing the association and dissociation rates of single Au(Ⅲ) ions under the applied voltages, it was found that a confined nanopore increased the bonding rate of Au(Ⅲ)-Histidine coordination reactions by around 105 times compared to the bulk solution, and the optimal voltage for single-molecule coordination., providing valuable insights for designing reaction pathways in electrochemical catalysis. This research revealed a novel mechanism for metal coordination and amino acid residues in protein nanoconfined space, highlighting the dynamic interactions between metal ions and amino acid residues and the importance of the confined effect, providing insights for developing efficient, eco-friendly electrocatalytic nanomaterials.","PeriodicalId":76,"journal":{"name":"Faraday Discussions","volume":"213 1","pages":""},"PeriodicalIF":3.4,"publicationDate":"2024-06-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141527892","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Adam Nelson, Wassilios Papawassiliou, Subhradip Paul, Sabine Hediger, Ivan Hung, Zhehong Gan, Amrit Venkatesh, W. Trent Trent Franks, Mark Edmund E Smith, David Gajan, Gaël De Paëpe, Christian Bonhomme, Danielle Laurencin, Christel Gervais
Octacalcium phosphate (OCP, Ca8(PO4)4(HPO4)2.5H2O) is a notable calcium phosphate due to its biocompatibility, making it a widely studied material for bone substitution. It is known to be a precursor of bone mineral, but its role in biomineralisation remains unclear. While the structure of OCP has been the subject of thorough investigations (including using Rietveld refinements of X-ray diffraction data, and NMR crystallography studies), important questions regarding the symmetry and H-bonding network in the material remain. In this study, it is shown that OCP undergoes a lowering of symmetry below 200 K, evidenced by 1H, 17O, 31P and 43Ca solid state NMR experiments. Using ab-initio molecular dynamics (MD) simulations and Gauge Including Projected Augmented Wave (GIPAW) DFT calculations of NMR parameters, the presence of rapid motion of the water molecules in the crystal cell at room temperature is proved. This information leads to an improved description of the OCP structure at both low and ambient temperatures, and helps explain long-standing issues of symmetry. Remaining challenges related to the understanding of the structure of OCP are then discussed.
磷酸八钙（OCP，Ca8(PO4)4(HPO4)2.5H2O）因其生物相容性而成为一种著名的磷酸钙，也因此成为一种被广泛研究的骨替代材料。众所周知，它是骨矿物质的前体，但其在生物矿化中的作用仍不清楚。虽然对 OCP 的结构进行了深入研究（包括使用 X 射线衍射数据的里特维尔德细化和核磁共振晶体学研究），但有关该材料的对称性和 H 键网络的重要问题仍然存在。本研究表明，OCP 在 200 K 以下会发生对称性降低的现象，1H、17O、31P 和 43Ca 固态核磁共振实验证明了这一点。利用非原位分子动力学（MD）模拟和 NMR 参数的量规包括投影增强波（GIPAW）DFT 计算，证明了在室温下晶胞中存在水分子的快速运动。这一信息改进了对 OCP 结构在低温和常温下的描述，并有助于解释长期存在的对称性问题。随后讨论了在理解 OCP 结构方面仍然存在的挑战。
{"title":"Temperature-induced mobility in Octacalcium Phosphate impacts crystal symmetry: water dynamics studied by NMR crystallography","authors":"Adam Nelson, Wassilios Papawassiliou, Subhradip Paul, Sabine Hediger, Ivan Hung, Zhehong Gan, Amrit Venkatesh, W. Trent Trent Franks, Mark Edmund E Smith, David Gajan, Gaël De Paëpe, Christian Bonhomme, Danielle Laurencin, Christel Gervais","doi":"10.1039/d4fd00108g","DOIUrl":"https://doi.org/10.1039/d4fd00108g","url":null,"abstract":"Octacalcium phosphate (OCP, Ca<small><sub>8</sub></small>(PO<small><sub>4</sub></small>)<small><sub>4</sub></small>(HPO<small><sub>4</sub></small>)<small><sub>2</sub></small>.5H<small><sub>2</sub></small>O) is a notable calcium phosphate due to its biocompatibility, making it a widely studied material for bone substitution. It is known to be a precursor of bone mineral, but its role in biomineralisation remains unclear. While the structure of OCP has been the subject of thorough investigations (including using Rietveld refinements of X-ray diffraction data, and NMR crystallography studies), important questions regarding the symmetry and H-bonding network in the material remain. In this study, it is shown that OCP undergoes a lowering of symmetry below 200 K, evidenced by <small><sup>1</sup></small>H, <small><sup>17</sup></small>O, <small><sup>31</sup></small>P and <small><sup>43</sup></small>Ca solid state NMR experiments. Using <em>ab-initio</em> molecular dynamics (MD) simulations and Gauge Including Projected Augmented Wave (GIPAW) DFT calculations of NMR parameters, the presence of rapid motion of the water molecules in the crystal cell at room temperature is proved. This information leads to an improved description of the OCP structure at both low and ambient temperatures, and helps explain long-standing issues of symmetry. Remaining challenges related to the understanding of the structure of OCP are then discussed.","PeriodicalId":76,"journal":{"name":"Faraday Discussions","volume":"27 1","pages":""},"PeriodicalIF":3.4,"publicationDate":"2024-06-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141527983","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The next generation of electroactive materials will depend on advanced nanomaterials, such as nanoparticles (NPs) for improved function and reduced cost. As such, the development of structure-function relationships for these NPs has become a prime focus for researchers from many fields, including materials science, catalysis, energy storage, photovoltaics, environmental/biomedical sensing, etc. The technique of scanning electrochemical cell microscopy (SECCM) has naturally positioned itself as a premier experimental methodology for the investigation of electroactive NPs, due to its unique capability to encapsulate individual, spatially distinct entities, and to apply a potential to (and measure the resulting current of) single-NPs. Over the course of conducting these single-NP investigations, a number of unexpected (i.e. rarely-reported) results have been collected, including fluctuating current responses, and carrying of the NP by the SECCM probe, hypothesised to be due to insufficient NP-surface interaction. Additionally, locations with measurable electrochemical activity have been found to contain no associated NP, and conversely locations with no activity have been found to contain NPs. Through presenting and discussing these findings, this article seeks to highlight the complications associated with single-NP SECCM measurements in order to endorse the broad inclusivity of data.
{"title":"Revealing the Diverse Electrochemistry of Nanoparticles with Scanning Electrochemical Cell Microscopy","authors":"Lachlan Gaudin, Cameron Luke Bentley","doi":"10.1039/d4fd00115j","DOIUrl":"https://doi.org/10.1039/d4fd00115j","url":null,"abstract":"The next generation of electroactive materials will depend on advanced nanomaterials, such as nanoparticles (NPs) for improved function and reduced cost. As such, the development of structure-function relationships for these NPs has become a prime focus for researchers from many fields, including materials science, catalysis, energy storage, photovoltaics, environmental/biomedical sensing, etc. The technique of scanning electrochemical cell microscopy (SECCM) has naturally positioned itself as a premier experimental methodology for the investigation of electroactive NPs, due to its unique capability to encapsulate individual, spatially distinct entities, and to apply a potential to (and measure the resulting current of) single-NPs. Over the course of conducting these single-NP investigations, a number of unexpected (i.e. rarely-reported) results have been collected, including fluctuating current responses, and carrying of the NP by the SECCM probe, hypothesised to be due to insufficient NP-surface interaction. Additionally, locations with measurable electrochemical activity have been found to contain no associated NP, and conversely locations with no activity have been found to contain NPs. Through presenting and discussing these findings, this article seeks to highlight the complications associated with single-NP SECCM measurements in order to endorse the broad inclusivity of data.","PeriodicalId":76,"journal":{"name":"Faraday Discussions","volume":"38 1","pages":""},"PeriodicalIF":3.4,"publicationDate":"2024-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141501896","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Proteins play crucial roles in nearly all biological activities, with their functional structures deriving from stable folded conformations. Protein denaturation, induced by chemical and physical agents, is a complex process where proteins lose their stable structures, thereby impairing their biological functions. Characterizing protein denaturation at the single-molecule level remains a significant challenge. In this study, we developed non-adhesive silicon nitride nanonets coated with polyethylene glycol to capture individual proteins. We utilized these nanonets to investigate the denaturation of ovalbumin induced by guanidine hydrochloride (Gdn-HCl) and lead chloride. The entire denaturation and renaturation processes of a single ovalbumin molecule were monitored via ionic current measurements through the nanonets. These non-sticky nanonets offer a versatile tool for real-time studies of structural changes during protein denaturation.
{"title":"Non-sticky SiNx nanonets for single protein denaturation analysis","authors":"Yuanhao Wang, Nan An, Bintong Huang, Yueming Zhai","doi":"10.1039/d4fd00117f","DOIUrl":"https://doi.org/10.1039/d4fd00117f","url":null,"abstract":"Proteins play crucial roles in nearly all biological activities, with their functional structures deriving from stable folded conformations. Protein denaturation, induced by chemical and physical agents, is a complex process where proteins lose their stable structures, thereby impairing their biological functions. Characterizing protein denaturation at the single-molecule level remains a significant challenge. In this study, we developed non-adhesive silicon nitride nanonets coated with polyethylene glycol to capture individual proteins. We utilized these nanonets to investigate the denaturation of ovalbumin induced by guanidine hydrochloride (Gdn-HCl) and lead chloride. The entire denaturation and renaturation processes of a single ovalbumin molecule were monitored via ionic current measurements through the nanonets. These non-sticky nanonets offer a versatile tool for real-time studies of structural changes during protein denaturation.","PeriodicalId":76,"journal":{"name":"Faraday Discussions","volume":"10 1","pages":""},"PeriodicalIF":3.4,"publicationDate":"2024-06-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141527985","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Data-driven techniques for establishing quantitative structure property relations are a pillar of modern materials and molecular discovery. Fuelled by the recent progress in deep learning methodology and the abundance of new algorithms, it is tempting to chase benchmarks and incrementally build ever more capable machine learning (ML) models. While model evaluation has made significant progress, the intrinsic limitations arising from the underlying experimental data are often overlooked. In the chemical sciences data collection is costly, thus datasets are small and experimental errors can be significant. These limitations of such datasets affect their predictive power, a fact that is rarely considered in a quantitative way. In this study, we analyse commonly used ML datasets for regression and classification from drug discovery, molecular discovery, and materials discovery. We derived maximum and realistic performance bounds for nine such datasets by introducing noise based on estimated or actual experimental errors. We then compared the estimated performance bounds to the reported performance of leading ML models in the literature. Out of the nine datasets and corresponding ML models considered, four were identified to have reached or surpassed dataset performance limitations and thus, they may potentially be fitting noise. More generally, we systematically examine how data range, the magnitude of experimental error, and the number of data points influence dataset performance bounds. Alongside this paper, we release the Python package NoiseEstimator and provide a web- based application for computing realistic performance bounds. This study and the resulting tools will help practitioners in the field understand the limitations of datasets and set realistic expectations for ML model performance. This work stands as a reference point, offering analysis and tools to guide development of future ML models in the chemical sciences.
建立定量结构属性关系的数据驱动技术是现代材料和分子发现的支柱。近年来，深度学习方法论取得了长足进步，新算法层出不穷，因此，追逐基准并逐步建立能力更强的机器学习（ML）模型很有诱惑力。虽然模型评估已经取得了重大进展，但底层实验数据带来的内在局限性往往被忽视。在化学科学领域，数据收集成本很高，因此数据集很小，实验误差可能很大。这些数据集的局限性影响了它们的预测能力，而这一事实却很少得到定量考虑。在本研究中，我们分析了药物发现、分子发现和材料发现中用于回归和分类的常用 ML 数据集。通过引入基于估计或实际实验误差的噪声，我们得出了九个此类数据集的最大和实际性能界限。然后，我们将估计的性能边界与文献中报道的主要 ML 模型的性能进行了比较。在考虑的九个数据集和相应的 ML 模型中，我们发现有四个已经达到或超过了数据集的性能限制，因此，它们有可能是拟合噪声。更广泛地说，我们系统地研究了数据范围、实验误差的大小和数据点的数量如何影响数据集的性能界限。在发表这篇论文的同时，我们还发布了 Python 软件包 NoiseEstimator，并提供了一个基于网络的应用程序，用于计算现实的性能边界。这项研究和由此产生的工具将帮助该领域的从业人员了解数据集的局限性，并对 ML 模型的性能设定切合实际的期望值。这项工作可作为一个参考点，为指导化学科学领域未来 ML 模型的开发提供分析和工具。
{"title":"Are we fitting data or noise? Analysing the predictive power of commonly used datasets in drug-, materials-, and molecular-discovery.","authors":"Daniel Crusius, Flaviu Cipcigan, Philip Biggin","doi":"10.1039/d4fd00091a","DOIUrl":"https://doi.org/10.1039/d4fd00091a","url":null,"abstract":"Data-driven techniques for establishing quantitative structure property relations are a pillar of modern materials and molecular discovery. Fuelled by the recent progress in deep learning methodology and the abundance of new algorithms, it is tempting to chase benchmarks and incrementally build ever more capable machine learning (ML) models. While model evaluation has made significant progress, the intrinsic limitations arising from the underlying experimental data are often overlooked. In the chemical sciences data collection is costly, thus datasets are small and experimental errors can be significant. These limitations of such datasets affect their predictive power, a fact that is rarely considered in a quantitative way. In this study, we analyse commonly used ML datasets for regression and classification from drug discovery, molecular discovery, and materials discovery. We derived maximum and realistic performance bounds for nine such datasets by introducing noise based on estimated or actual experimental errors. We then compared the estimated performance bounds to the reported performance of leading ML models in the literature. Out of the nine datasets and corresponding ML models considered, four were identified to have reached or surpassed dataset performance limitations and thus, they may potentially be fitting noise. More generally, we systematically examine how data range, the magnitude of experimental error, and the number of data points influence dataset performance bounds. Alongside this paper, we release the Python package NoiseEstimator and provide a web- based application for computing realistic performance bounds. This study and the resulting tools will help practitioners in the field understand the limitations of datasets and set realistic expectations for ML model performance. This work stands as a reference point, offering analysis and tools to guide development of future ML models in the chemical sciences.","PeriodicalId":76,"journal":{"name":"Faraday Discussions","volume":"25 1","pages":""},"PeriodicalIF":3.4,"publicationDate":"2024-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141256072","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Taiana L.E. Pereira, Jon Serrano-Sevillano, Beatriz Diaz Moreno, Joel Reid, Dany Carlier, Gillian Goward
In our recent study, we demonstrated using 7Li solid-state Nuclear Magnetic Resonance (ssNMR) and single-crystal X-ray diffraction, that the cathode LiFeV2O7 possesses a defect associated with the positioning of vanadium atoms. We proposed that this defect could be the source of extra signals detected in the 7Li NMR spectra. In this context, we now apply density functional theory (DFT) calculations to assign the experimental signals observed in 7Li NMR spectra of the pristine sample. The calculation results are in strong agreement with the experimental observations. DFT calculations are a useful tool to interpret the observed paramagnetic shifts and understand how the presence of disorder affects the spectra behavior through the spin-density transfer processes. Furthermore, we conducted a detailed study of the lithiated phase combining operando synchrotron powder X-ray diffraction (SPXRD) and DFT calculations. A noticeable volume expansion is observed through the first discharge cycle which likely contributes to the enhanced lithium dynamics in the bulk material, as supported by previously published ssNMR data. DFT calculations are used to model the lithiated phase and demonstrate that both iron and vanadium participate in the redox process. The unusual electronic structure of the V4+ -exhibits a single electron on the 3dxy orbital perpendicular to the V-O-Li bond being a source of a negative Fermi contact shift observed in the 7Li NMR of the lithiated phase.
{"title":"A Combined 7Li NMR, Density Functional Theory and Operando Synchrotron X-Ray Powder Diffraction to Investigate a Structural Evolution of Cathode Material LiFeV2O7","authors":"Taiana L.E. Pereira, Jon Serrano-Sevillano, Beatriz Diaz Moreno, Joel Reid, Dany Carlier, Gillian Goward","doi":"10.1039/d4fd00077c","DOIUrl":"https://doi.org/10.1039/d4fd00077c","url":null,"abstract":"In our recent study, we demonstrated using <small><sup>7</sup></small>Li solid-state Nuclear Magnetic Resonance (ssNMR) and single-crystal X-ray diffraction, that the cathode LiFeV<small><sub>2</sub></small>O<small><sub>7</sub></small> possesses a defect associated with the positioning of vanadium atoms. We proposed that this defect could be the source of extra signals detected in the <small><sup>7</sup></small>Li NMR spectra. In this context, we now apply density functional theory (DFT) calculations to assign the experimental signals observed in 7Li NMR spectra of the pristine sample. The calculation results are in strong agreement with the experimental observations. DFT calculations are a useful tool to interpret the observed paramagnetic shifts and understand how the presence of disorder affects the spectra behavior through the spin-density transfer processes. Furthermore, we conducted a detailed study of the lithiated phase combining operando synchrotron powder X-ray diffraction (SPXRD) and DFT calculations. A noticeable volume expansion is observed through the first discharge cycle which likely contributes to the enhanced lithium dynamics in the bulk material, as supported by previously published ssNMR data. DFT calculations are used to model the lithiated phase and demonstrate that both iron and vanadium participate in the redox process. The unusual electronic structure of the V<small><sup>4+</sup></small> -exhibits a single electron on the 3d<small><sub>xy</sub></small> orbital perpendicular to the V-O-Li bond being a source of a negative Fermi contact shift observed in the <small><sup>7</sup></small>Li NMR of the lithiated phase.","PeriodicalId":76,"journal":{"name":"Faraday Discussions","volume":"98 1","pages":""},"PeriodicalIF":3.4,"publicationDate":"2024-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141256069","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Christopher Taylor, Patrick Butler, Graeme Matthew Day
Computational crystal structure prediction (CSP) is an increasingly powerful technique in materials discovery, due to its ability to reveal trends and permit insight across the possibility space of crystal structures of a candidate molecule, beyond simply the observed structure(s). In this work, we demonstrate the reliability and scalability of CSP methods for small, rigid organic molecules by performing in-depth CSP investigations for over 1000 such compounds, the largest survey of its kind to-date. We show that this highly-efficient force-field-based CSP approach is superbly predictive, locating 99.4% of observed experimental structures, and ranking a large majority of these (74%) as among the most stable possible structures (to within uncertainty due to thermal effects). We present two examples of insights such large predicted datasets can permit, examining the space group preferences of organic molecular crystals and rationalising empirical rules concerning the spontaneous resolution of chiral molecules. Finally, we exploit this large and diverse dataset for developing transferable machine-learned energy potentials for the organic solid state, training a neural network lattice energy correction to force field energies that offers substantial improvements to the already impressive energy rankings, and a MACE equivariant message-passing neural network for crystal structure reoptimisation. We conclude that the excellent performance and reliability of the CSP workflow enables the creation of very large datasets of broad utility and explanatory power in materials design.
{"title":"Predictive crystallography at scale: mapping, validating, and learning from 1,000 crystal energy landscapes","authors":"Christopher Taylor, Patrick Butler, Graeme Matthew Day","doi":"10.1039/d4fd00105b","DOIUrl":"https://doi.org/10.1039/d4fd00105b","url":null,"abstract":"Computational crystal structure prediction (CSP) is an increasingly powerful technique in materials discovery, due to its ability to reveal trends and permit insight across the possibility space of crystal structures of a candidate molecule, beyond simply the observed structure(s). In this work, we demonstrate the reliability and scalability of CSP methods for small, rigid organic molecules by performing in-depth CSP investigations for over 1000 such compounds, the largest survey of its kind to-date. We show that this highly-efficient force-field-based CSP approach is superbly predictive, locating 99.4% of observed experimental structures, and ranking a large majority of these (74%) as among the most stable possible structures (to within uncertainty due to thermal effects). We present two examples of insights such large predicted datasets can permit, examining the space group preferences of organic molecular crystals and rationalising empirical rules concerning the spontaneous resolution of chiral molecules. Finally, we exploit this large and diverse dataset for developing transferable machine-learned energy potentials for the organic solid state, training a neural network lattice energy correction to force field energies that offers substantial improvements to the already impressive energy rankings, and a MACE equivariant message-passing neural network for crystal structure reoptimisation. We conclude that the excellent performance and reliability of the CSP workflow enables the creation of very large datasets of broad utility and explanatory power in materials design.","PeriodicalId":76,"journal":{"name":"Faraday Discussions","volume":"36 1","pages":""},"PeriodicalIF":3.4,"publicationDate":"2024-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141256188","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Lei Chen, Carlos Bornes, Oscar Bengtsson, Andreas Erlebach, Ben Slater, Lukáš Grajciar, Christopher J. Heard
One of the main limitations in supporting experimental characterization of Al siting/pairing via modelling is the high computational cost of ab initio calculations. For this reason, most works rely on static or very short dynamical simulations, considering limited Al pairing/siting combinations. As a result, comparison with experiment suffers from a large degree of uncertainty. To alleviate this limitation we have developed neural network potentials (NNPs) which can dynamically sample across broad configurational and chemical spaces of sodium-form aluminosilicate zeolites, preserving the level of accuracy of the ab initio (dispersion-corrected metaGGA) training set. By exploring a wide range of Al/Na arrangements and a combination of experimentally relevant Si/Al ratios, we found that the 23Na NMR spectra of dehydrated high-silica CHA zeolite offer an opportunity to assess the distribution and pairing of Al atoms. We observed that the 23Na chemical shift is sensitive not only to the location of sodium in 6- and 8MRs, but also to the Al-Sin-Al sequence length. Furthermore, neglect of thermal and dynamical contributions were found to lead to errors of several ppm, and have a profound influence on the shape of the spectra and the dipolar coupling constants, thus necessitating the long-term dynamical simulations made feasible by NNPs. Finally, we obtained a predictive regression model for 23Na chemical shift in CHA (Si/Al = 35, 17, 11) that circumvents the need for expensive NMR density functional calculations and can be easily extended to other zeolite frameworks. By combining NNPs and regression methods, we can expedite the simulations of NMR properties and capture the effect dynamics on the spectra, which is often overlooked in computational studies despite its clear manifestation in experimental setups.
{"title":"A machine learning approach for dynamical modelling of Al distributions in zeolites via 23Na/27Al solid-state NMR","authors":"Lei Chen, Carlos Bornes, Oscar Bengtsson, Andreas Erlebach, Ben Slater, Lukáš Grajciar, Christopher J. Heard","doi":"10.1039/d4fd00100a","DOIUrl":"https://doi.org/10.1039/d4fd00100a","url":null,"abstract":"One of the main limitations in supporting experimental characterization of Al siting/pairing via modelling is the high computational cost of ab initio calculations. For this reason, most works rely on static or very short dynamical simulations, considering limited Al pairing/siting combinations. As a result, comparison with experiment suffers from a large degree of uncertainty. To alleviate this limitation we have developed neural network potentials (NNPs) which can dynamically sample across broad configurational and chemical spaces of sodium-form aluminosilicate zeolites, preserving the level of accuracy of the ab initio (dispersion-corrected metaGGA) training set. By exploring a wide range of Al/Na arrangements and a combination of experimentally relevant Si/Al ratios, we found that the <small><sup>23</sup></small>Na NMR spectra of dehydrated high-silica CHA zeolite offer an opportunity to assess the distribution and pairing of Al atoms. We observed that the <small><sup>23</sup></small>Na chemical shift is sensitive not only to the location of sodium in 6- and 8MRs, but also to the Al-Si<small><sub>n</sub></small>-Al sequence length. Furthermore, neglect of thermal and dynamical contributions were found to lead to errors of several ppm, and have a profound influence on the shape of the spectra and the dipolar coupling constants, thus necessitating the long-term dynamical simulations made feasible by NNPs. Finally, we obtained a predictive regression model for <small><sup>23</sup></small>Na chemical shift in CHA (Si/Al = 35, 17, 11) that circumvents the need for expensive NMR density functional calculations and can be easily extended to other zeolite frameworks. By combining NNPs and regression methods, we can expedite the simulations of NMR properties and capture the effect dynamics on the spectra, which is often overlooked in computational studies despite its clear manifestation in experimental setups.","PeriodicalId":76,"journal":{"name":"Faraday Discussions","volume":"52 1","pages":""},"PeriodicalIF":3.4,"publicationDate":"2024-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141256526","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：