Cambridge Structural Database Research Articles

Accelerating material property prediction using generically complete isometry invariants

Periodic material or crystal property prediction using machine learning has grown popular in recent years as it provides a computationally efficient replacement for classical simulation methods. A crucial first step for any of these algorithms is the representation used for a periodic crystal. While similar objects like molecules and proteins have a finite number of atoms and their representation can be built based upon a finite point cloud interpretation, periodic crystals are unbounded in size, making their representation more challenging. In the present work, we adapt the Pointwise Distance Distribution (PDD), a continuous and generically complete isometry invariant for periodic point sets, as a representation for our learning algorithm. The PDD distinguished all (more than 660 thousand) periodic crystals in the Cambridge Structural Database as purely periodic sets of points without atomic types. We develop a transformer model with a modified self-attention mechanism that combines PDD with compositional information via a spatial encoding method. This model is tested on the crystals of the Materials Project and Jarvis-DFT databases and shown to produce accuracy on par with state-of-the-art methods while being several times faster in both training and prediction time.

DMXP: A De Novo Small-Molecule 3D Structure Predictor with Graph Attention Networks.

Generating the three-dimensional (3D) structure of small molecules is crucial in both structure- and ligand-based drug design. Structure-based drug design needs bioactive conformations of compounds for lead identification and optimization. Ligand-based drug design techniques, such as 3D shape similarity search, 3D pharmacophore model, 3D-QSAR, etc., all require high-quality small-molecule ligand conformations to obtain reliable results. Although predicting a small molecular bioactive conformer requires information from the receptor, a crystal structure of the molecule is a proper approximation to its bioactive conformer in a specific receptor because the binding pose of a small molecule in its receptor's binding pockets should be energetically close to the crystal structures. This study presents a de novo small molecular structure predictor (dMXP) with graph attention networks based on crystal data derived from the Cambridge Structural Database (CSD) combined with molecular electrostatic information calculated by density-functional theory (DFT). Two featuring strategies (topological and atomic partial change features) were employed to explore the relation between these features and the 3D crystal structure of a small molecule. These features were then assembled to construct the holistic 3D crystal structure of a molecule. Molecular graphs were encoded using a graph attention mechanism to deal with the issues of the inconsistencies of local substructures contributing to the entire molecular structure. The root-mean-square deviation (RMSDs) of approximately 80% dMXP predicted structures and the native binding poses within receptors are less than 2.0 Å.

Time-series analysis of rhenium(I) organometallic covalent binding to a model protein for drug development.

Metal-based complexes with their unique chemical properties, including multiple oxidation states, radio-nuclear capabilities and various coordination geometries yield value as potential pharmaceuticals. Understanding the interactions between metals and biological systems will prove key for site-specific coordination of new metal-based lead compounds. This study merges the concepts of target coordination with fragment-based drug methodologies, supported by varying the anomalous scattering of rhenium along with infrared spectroscopy, and has identified rhenium metal sites bound covalently with two amino acid types within the model protein. A time-based series of lysozyme-rhenium-imidazole (HEWL-Re-Imi) crystals was analysed systematically over a span of 38 weeks. The main rhenium covalent coordination is observed at His15, Asp101 and Asp119. Weak (i.e. noncovalent) interactions are observed at other aspartic, asparagine, proline, tyrosine and tryptophan side chains. Detailed bond distance comparisons, including precision estimates, are reported, utilizing the diffraction precision index supplemented with small-molecule data from the Cambridge Structural Database. Key findings include changes in the protein structure induced at the rhenium metal binding site, not observed in similar metal-free structures. The binding sites are typically found along the solvent-channel-accessible protein surface. The three primary covalent metal binding sites are consistent throughout the time series, whereas binding to neighbouring amino acid residues changes through the time series. Co-crystallization was used, consistently yielding crystals four days after setup. After crystal formation, soaking of the compound into the crystal over 38 weeks is continued and explains these structural adjustments. It is the covalent bond stability at the three sites, their proximity to the solvent channel and the movement of residues to accommodate the metal that are important, and may prove useful for future radiopharmaceutical development including target modification.

Material Property Prediction Using Graphs Based on Generically Complete Isometry Invariants

AbstractThe structure–property hypothesis says that the properties of all materials are determined by an underlying crystal structure. The main obstacle was the ambiguity of conventional crystal representations based on incomplete or discontinuous descriptors that allow false negatives or false positives. This ambiguity was resolved by the ultra-fast pointwise distance distribution, which distinguished all periodic structures in the world’s largest collection of real materials (Cambridge structural database). State-of-the-art results in property prediction were previously achieved by graph neural networks based on various graph representations of periodic crystals, including the Crystal Graph with vertices at all atoms in a crystal unit cell. This work adapts the pointwise distance distribution for a simpler graph whose vertex set is not larger than the asymmetric unit of a crystal structure. The new Distribution Graph reduces mean absolute error by 0.6–12% while having 44–88% of the number of vertices when compared to the Crystal Graph when applied on the Materials Project and Jarvis-DFT datasets using CGCNN and ALIGNN. Methods for hyper-parameters selection for the graph are backed by the theoretical results of the pointwise distance distribution and are then experimentally justified.

Does the Central Nitrogen Atom Make a Difference? A Comparison of Non‐Coordinating Pyridine and Benzene Spacers in Multitopic Ligands

AbstractIn metal coordination compounds of the divergent ligands 4,2′:6′,4“‐terpyridine and 3,2′:6′,3”‐terpyridine and their 4′‐functionalized derivatives, the central pyridine ring of the 4,2′:6′,4“‐terpyridine and 3,2′:6′,3”‐terpyridine domains is non‐coordinated. We present an overview of data from the Cambridge Structural Database to assess whether there are structural similarities between the coordination compounds of 4,2′:6′,4“‐tpy and 1,3‐di(pyridin‐4‐yl)benzene, and between 3,2′:6′,3”‐tpy and 1,3‐di(pyridin‐3‐yl)benzene. Based upon structurally characterized compounds, it emerges that the coordination chemistry of ligands including one or more 4,2′:6′,4“‐terpyridine or 3,2′:6′,3”‐terpyridine metal‐binding domains is more abundantly exemplified than that of corresponding ligands based upon 1,3‐di(pyridin‐4‐yl)benzene and 1,3‐di(pyridin‐3‐yl)benzene. We provide an overview of metallamacrocycles and cages, 1D‐coordination polymers and 2D‐ and 3D‐networks.

Pyramidalization of the Carboxamide sp2-Center in Peptide Structures.

A CSD search in the Cambridge Crystallographic Database for the substructure N-CαH-C'(═O)-N gave 24,180 peptide structures for analysis of the pyramidalization of the sp2-hybridized carboxamide group C'(═O)NCα, which had not been investigated before. The dependence of the pyramidalization θ = O-N-C'-Cα on the rotation angle ψ = O═C'-Cα-N about bond C'-Cα resulted in a curve with three maxima, three minima, and six zero-crossings. Surprisingly, the ψ/θ analysis of the individual amino acid building blocks showed that all of them exhibited similar curves, irrespective of their different R substituents. This unusual behavior is explained by a 3-fold short-range potential set up by the three covalent bonds, emanating from Cα. The tie-up of the rotation angle ψ and the pyramidalization θ in a rigid coupling is remarkable. In the 24,180 peptide structures, subjected to X-ray crystallography, there is no dynamics. For peptides in solution, the rotation/pyramidalization curve ψ/θav determines the degree of pyramidalization θ, when the rotation angle ψ runs through a full 360° circle. Density functional theory (DFT) calculations of alaninamide supported the analysis.

Structure and Bonding in π-Stacked Perylenes: The Impact of Charge on Pancake Bonding.

Perylene (PER) is a prototype of polycyclic aromatic hydrocarbons (PAHs), which play a pivotal role in various functional and electronic materials due to favorable molecule-to-molecule overlaps, which enhance electronic transport. This study provides guidelines regarding the impact of molecular charge on pancake bonding, a form of strong π-stacking interaction. Pancake bonding significantly boosts interaction energies within the monopositive dimer ([(C20H12)2]•+ or PER2+), crucial for stabilizing aggregation and crystal formation. We discovered energetically feasible sliding and rotation pathways within the [(C20H12)2]•+ dimer, connecting different configurations found in the Cambridge Structural Database (CSD). The dimer's charge profoundly influences the pancake bond order (PBO) and the strength and structural preferences of pancake bonding. The most stable configuration is found in the monocationic state (PER2+), featuring a pancake bond order of 1/2 with one-electron multicenter bonding (1e/mc) with similar characteristics for charge -1. Increasing the total charge of the dimer to +2 or -2 leads to an unstable local minimum. Diverse distribution of pancake bonding types present in crystal structures is interpreted with modeling based on dimer computations with varying charges.

Structural Analysis of 3,5-Bistrifluoromethylhydrocinnamic Acid

The crystal structure of 3,5-bistrifluoromethylhydrocinnamic acid [systematic name: 3-[3,5-bis(trifluoromethyl)phenyl]propanoic acid], C11H8F6O2, has been determined and described. The structure was subject to the Hirshfeld surface-analysis and CE-B3LYP interaction-energies calculations. The title compound crystallises in the monoclinic P21/c space group with one molecule in the asymmetric unit. The propanoic acid side chain of the studied molecule has a bent conformation. The key supramolecular motif in the crystal structure is a centrosymmetric O–H∙∙∙O hydrogen-bonded dimer (R22(8) in the graph set notation). According to CE-B3LYP, the molecules involved in this motif exhibit the strongest pairwise interaction total energy (Etot = −67.9 kJ/mol). On the other hand, there are seven other interacting molecular pairs with significant Etot values in the range of −17 to −28 kJ/mol. In these, the energy is dominated by the dispersive contribution. A survey of the Cambridge Structural Database revealed that in other 3-phenylpropanoic acid structures, the middle dihedral angle of the propanoic acid side chain is always in the trans conformation. This contrasts the current structure where this dihedral angle is in the gauche conformation. According to the Density Functional Theory calculations in the gas phase (at the B3LYP/aug-cc-pvDZ level), the presence of the two CF3 groups (strong electron-withdrawing character) increases the population of the gauche conformers by a substituent electronic effect, and this may be a minor factor contributing to the appearance of this conformation observed in the solid state.

Cocrystal Prediction Based on Deep Forest Model—A Case Study of Febuxostat

To aid cocrystal screening, a deep forest-based cocrystal prediction model was developed in this study using data from the Cambridge Structural Database (CSD). The positive samples in the experiment came from the CSD. The negative samples were partly from the failure records in other papers, and some were randomly generated according to specific rules, resulting in a total of 8576 pairs. Compared with the models of traditional machine learning methods and simple deep neural networks models, the deep forest model has better performance and faster training speed. The accuracy is about 95% on the test set. Febuxostat cocrystal screening was also tested to verify the validity of the model. Our model correctly predicted the formation of cocrystal. It shows that our model is practically useful in practice.

Elucidating the Polymorphism of Xanthone: A Crystallization and Characterization Study.

The aim of this work is to shed light on the polymorphism of xanthones, a class of oxygenated molecules well known for their bioactivity, including antioxidant, anticancer, and anti-inflammatory effects. Understanding the polymorphism of xanthones can enable the design of novel solid products for pharmaceutical, nutraceutical, and agrochemical applications. Prior to this work, two entries accounting for different space groups were deposited for 9-xanthone in the Cambridge Structure Database (CSD): an orthorhombic P212121 and a monoclinic P21 structure solved at room and low temperatures, respectively. However, the very high similarity between these two structures and the lack of clear differences in their physical properties (e.g., thermal behavior) suggested the possibility of the existence of only one crystal structure. In fact, the differences shown in the literature data might be related to the chosen operating parameters, as well as the instrumental resolution of the single-crystal X-ray diffraction experiments. In the work presented here, the ambiguity in the polymorphism of xanthone is investigated using thermal analysis, powder and synchrotron single-crystal XRD, and optical microscopy. Additionally, a workflow for the correct identification of twinned crystal structures, which can be applied to other polymorphic systems, is presented. Such workflow combines the collection of a large data set of high-resolution diffraction patterns using synchrotron radiation with the use of principal component analysis, a dimensionality reduction technique, for a quick and effective identification of phase transitions happening during the data collection. Crystallization experiments were designed to promote the formation of different crystal structures of xanthone that were recrystallized based on past literature and beyond.

Predicting Lewis Acidity: Machine Learning the Fluoride Ion Affinity of p‐Block‐Atom‐Based Molecules

Abstract“How strong is this Lewis acid?” is a question researchers often approach by calculating its fluoride ion affinity (FIA) with quantum chemistry. Here, we present FIA49k, an extensive FIA dataset with 48,986 data points calculated at the RI‐DSD‐BLYP‐D3(BJ)/def2‐QZVPP//PBEh‐3c level of theory, including 13 different p‐block atoms as the fluoride accepting site. The FIA49k dataset was used to train FIA‐GNN, two message‐passing graph neural networks, which predict gas and solution phase FIA values of molecules excluded from training with a mean absolute error of 14 kJ mol−1 (r2=0.93) from the SMILES string of the Lewis acid as the only input. The level of accuracy is notable, given the wide energetic range of 750 kJ mol−1 spanned by FIA49k. The model's value was demonstrated with four case studies, including predictions for molecules extracted from the Cambridge Structural Database and by reproducing results from catalysis research available in the literature. Weaknesses of the model are evaluated and interpreted chemically. FIA‐GNN and the FIA49k dataset can be reached via a free web app (www.grebgroup.de/fia‐gnn).

Predicting Lewis Acidity: Machine Learning the Fluoride Ion Affinity of p-Block-Atom-Based Molecules.

"How strong is this Lewis acid?" is a question researchers often approach by calculating its fluoride ion affinity (FIA) with quantum chemistry. Here, we present FIA49k, an extensive FIA dataset with 48,986 data points calculated at the RI-DSD-BLYP-D3(BJ)/def2-QZVPP//PBEh-3c level of theory, including 13 different p-block atoms as the fluoride accepting site. The FIA49k dataset was used to train FIA-GNN, two message-passing graph neural networks, which predict gas and solution phase FIA values of molecules excluded from training with a mean absolute error of 14 kJ mol-1 (r2 =0.93) from the SMILES string of the Lewis acid as the only input. The level of accuracy is notable, given the wide energetic range of 750 kJ mol-1 spanned by FIA49k. The model's value was demonstrated with four case studies, including predictions for molecules extracted from the Cambridge Structural Database and by reproducing results from catalysis research available in the literature. Weaknesses of the model are evaluated and interpreted chemically. FIA-GNN and the FIA49k dataset can be reached via a free web app (www.grebgroup.de/fia-gnn).

Physics-Informed Generative Model for Drug-like Molecule Conformers.

We present a diffusion-based generative model for conformer generation. Our model is focused on the reproduction of the bonded structure and is constructed from the associated terms traditionally found in classical force fields to ensure a physically relevant representation. Techniques in deep learning are used to infer atom typing and geometric parameters from a training set. Conformer sampling is achieved by taking advantage of recent advancements in diffusion-based generation. By training on large, synthetic data sets of diverse, drug-like molecules optimized with the semiempirical GFN2-xTB method, high accuracy is achieved for bonded parameters, exceeding that of conventional, knowledge-based methods. Results are also compared to experimental structures from the Protein Databank and the Cambridge Structural Database.

Crystal structure of tetra-phenyl phosphate tetra-kis-[dimethyl (2,2,2-tri-chloro-acet-yl)phos-pho-ramidato]lutetium(III), PPh4[LuL4].

A lutetium(III) complex based on the anion of the ligand dimethyl (2,2,2-tri-chloro-acet-yl)phospho-ramidate (HL) and tetra-phenylphosphonium, of composition PPh4[LuL 4] (L = CAPh = carbacyl-amido-phosphate), or (C24H20)[Lu(C4H6Cl3NO4P)4], has been synthesized and structurally characterized. The X-ray diffraction study of the compound revealed that the lutetium ion is surrounded by four bis-chelating CAPh ligands, forming the complex anion [LuL 4]- with a coordination number of 8[O] for LuIII, while PPh4 + serves as a counter-ion. The coordination geometry around the Lu3+ ion was determined to be a nearly perfect triangular dodeca-hedron. The complex crystallizes in the monoclinic crystal system, space group P21/c, with four mol-ecules in the unit cell. Weak hydrogen bonds O⋯HC(Ph), Cl⋯HC(Ph) and N⋯HC(Ph) are formed between the cations and anions. For a comparative study, HL-based structures were retrieved from the Cambridge Structural Database (CSD) and their geometries and conformations are discussed. A Hirshfeld surface analysis was also performed.

The Cambridge Structural Database and structural dynamics.

With the availability of the computer readable information in the Cambridge Structural Database (CSD), wide ranging, largely automated comparisons of fragment, molecular, and crystal structures have become possible. They show that the distributions of interatomic distances, angles, and torsion angles for a given structural fragment occurring in different environments are highly correlated among themselves and with other observables such as spectroscopic signals, reaction and activation energies. The correlations often extend continuously over large ranges of parameter values. They are reminiscent of bond breaking and forming reactions, polyhedral rearrangements, and conformational changes. They map-qualitatively-the regions of the structural parameter space in which molecular dynamics take place, namely, the low energy regions of the respective (free) energy surfaces. The extension and continuous nature of the correlations provides an organizing principle of large groups of structural data and suggests a reconsideration of traditional definitions and descriptions of bonds, "nonbonded" and "noncovalent" interactions in terms of Lewis acids interacting with Lewis bases. These aspects are illustrated with selected examples of historic importance and with some later developments. It seems that the amount of information in the CSD (and other structural databases) and the knowledge on the nature of, and the correlations within, this body of information should allow one-in the near future-to make credible interpolations and possibly predictions of structures and their properties with machine learning methods.

Polymorph Identification for Flexible Molecules: Linear Regression Analysis of Experimental and Calculated Solution- and Solid-State NMR Data.

The Δδ regression approach of Blade et al. [ J. Phys. Chem. A 2020, 124(43), 8959-8977] for accurately discriminating between solid forms using a combination of experimental solution- and solid-state NMR data with density functional theory (DFT) calculation is here extended to molecules with multiple conformational degrees of freedom, using furosemide polymorphs as an exemplar. As before, the differences in measured 1H and 13C chemical shifts between solution-state NMR and solid-state magic-angle spinning (MAS) NMR (Δδexperimental) are compared to those determined by gauge-including projector augmented wave (GIPAW) calculations (Δδcalculated) by regression analysis and a t-test, allowing the correct furosemide polymorph to be precisely identified. Monte Carlo random sampling is used to calculate solution-state NMR chemical shifts, reducing computation times by avoiding the need to systematically sample the multidimensional conformational landscape that furosemide occupies in solution. The solvent conditions should be chosen to match the molecule's charge state between the solution and solid states. The Δδ regression approach indicates whether or not correlations between Δδexperimental and Δδcalculated are statistically significant; the approach is differently sensitive to the popular root mean squared error (RMSE) method, being shown to exhibit a much greater dynamic range. An alternative method for estimating solution-state NMR chemical shifts by approximating the measured solution-state dynamic 3D behavior with an ensemble of 54 furosemide crystal structures (polymorphs and cocrystals) from the Cambridge Structural Database (CSD) was also successful in this case, suggesting new avenues for this method that may overcome its current dependency on the prior determination of solution dynamic 3D structures.

Glycerol as Ligand in Metal Complexes—A Structural Review

The molecule glycerol (H3gly) plays a key role in sustainable and green chemistry. Having been discovered for over 200 years and produced from vegetable oils and animal fats by hydrolysis, saponification and transesterification reactions, this natural triol is today employed in a wide range of cosmetic, food, polymer and pharmaceutical applications. Moreover, it is an essential C3 precursor in the chemical industry, used in the production of several intermediates and it avoids the need for petroleum-based precursors. Less famous but just as exciting, in the domain of coordination chemistry, glycerol is also proving to be a suitable ligand, capable of binding to one or more metal centres, either directly in its triol H3gly form (rather rare), or in its various deprotonated glycerolate forms, such as [H2gly]−, [Hgly]2− and [gly]3− (in most cases). Since the 1970s, various molecular structures prepared from glycerol and metallic and organometallic precursors, ranging from mononuclear complexes to sophisticated aggregates and coordination polymers, have been isolated and characterised. On the basis of the single-crystal X-ray diffraction structures described so far in the literature and deposited in the Cambridge Structural Database, in this structural inventory, we review the different modes of coordination of glycerol and glycerolates with metals.

MACE: Automated Assessment of Stereochemistry of Transition Metal Complexes and Its Applications in Computational Catalysis.

Computational chemistry pipelines typically commence with geometry generation, well-established for organic compounds but presenting a considerable challenge for transition metal complexes. This paper introduces MACE, an automated computational workflow for converting chemist SMILES/MOL representations of the ligands and the metal center to 3D coordinates for all feasible stereochemical configurations for mononuclear octahedral and square planar complexes directly suitable for quantum chemical computations and implementation in high-throughput computational chemistry workflows. The workflow is validated through a structural screening of a data set of transition metal complexes extracted from the Cambridge Structural Database. To further illustrate the power and capabilities of MACE, we present the results of a model DFT study on the hemilability of pincer ligands in Ru, Fe, and Mn complexes, which highlights the utility of the workflow for both focused mechanistic studies and larger-scale high-throughput pipelines.

A plant cyclic nonapeptide of orbitide type: an electron density study

Abstract The electron density distribution (EDD) of a plant cyclic nonapeptide of orbitide type was studied. Crystal X-ray diffraction data was obtained from the Cambridge Structural Database (CSD) and refitted using scattering factors of the invariom library, thereby providing aspherical electron density. Bond topological, atomic properties and molecular surfaces (electrostatic potential and Hirshfeld surfaces) were derived. The partial double bond character of the peptide bond was confirmed by the ellipticity ε = 0.25. For eight N–H⋯O hydrogen bonds, atomic charges of contributing atoms differ depending on the type of the accepting oxygen atom. Atomic charge differences between negative main and positive side chains of this nonapeptide result in characteristic features of the electrostatic potential, which shows a positive isosurface around the molecule leading to repulsive interactions in the solid state structure. Weak intermolecular interactions are indicated by insignificant ED concentrations on the Hirshfeld surface except for weak signals at sites of intermolecular N–H⋯O and C–H⋯O hydrogen bonds.

The prediction of single-molecule magnet properties via deep learning.

This paper uses deep learning to present a proof-of-concept for data-driven chemistry in single-molecule magnets (SMMs). Previous discussions within SMM research have proposed links between molecular structures (crystal structures) and single-molecule magnetic properties; however, these have only interpreted the results. Therefore, this study introduces a data-driven approach to predict the properties of SMM structures using deep learning. The deep-learning model learns the structural features of the SMM molecules by extracting the single-molecule magnetic properties from the 3D coordinates presented in this paper. The model accurately determined whether a molecule was a single-molecule magnet, with an accuracy rate of approximately 70% in predicting the SMM properties. The deep-learning model found SMMs from 20 000 metal complexes extracted from the Cambridge Structural Database. Using deep-learning models for predicting SMM properties and guiding the design of novel molecules is promising.