The DP5 probability, quantification and visualisation of structural uncertainty in single molecules

Experimental and Computational NMR Studies of Large Alkaloids Exemplified With Vindoline Trimer: Advantages and Limitations

The complete 1H and 13C NMR assignments of a trimeric vindoline together with its individual components, dimeric vindolicine and monomeric vindoline, are performed based on a thorough analysis of the ROESY, COSY, HSQC, and HMBC spectra in combination with the state-of-the-art quantum-chemical calculations. A spatial structure of vindoline trimer is determined by means of computational conformational analysis in combination with the probability distribution map of its basic conformers. On the example of monoterpene indole alkaloid, the trimer vindoline, the present study reveals the power of modern computational NMR to perform identification and stereochemical studies of large natural compounds with some limitations, which may arise in the quantum chemical computing workflow.

Uncertainty Qualification for Deep Learning-Based Elementary Reaction Property Prediction

The prediction of the thermodynamic and kinetic properties of elementary reactions has shown rapid improvement due to the implementation of deep learning (DL) methods. While various studies have reported the success in predicting reaction properties, the quantification of prediction uncertainty has seldom been investigated, thus compromising the confidence in using these predicted properties in practical applications. Here, we integrated graph convolutional neural networks (GCNN) with three uncertainty prediction techniques, including deep ensemble, Monte Carlo (MC)-dropout, and evidential learning, to provide insights into the uncertainty quantification and utility. The deep ensemble model outperforms others in accuracy and shows the highest reliability in estimating prediction uncertainty across all elementary reaction property data sets. We also verified that the deep ensemble model showed a satisfactory capability in recognizing epistemic and aleatoric uncertainties. Additionally, we adopted a Monte Carlo Tree Search method for extracting the explainable reaction substructures, providing a chemical explanation for DL predicted properties and corresponding uncertainties. Finally, to demonstrate the utility of uncertainty qualification in practical applications, we performed an uncertainty-guided calibration of the DL-constructed kinetic model, which achieved a 25% higher hit ratio in identifying dominant reaction pathways compared to that of the calibration without uncertainty guidance.

Chemical shift prediction in 13C NMR spectroscopy using ensembles of message passing neural networks (MPNNs)

This study reports a deep learning approach that utilises message passing neural networks (MPNNs) for predicting chemical shifts in 13C NMR spectra of small molecules. MPNNs were trained on two distinct datasets: one with approximately 4000 labelled structures and another with over 40,000. To reduce stochastic variation, an ensemble framework was implemented, which is simple to deploy on multiple nodes of a High-Performance Computing facility. The results emphasise the critical role of training set size and diversity. While prediction performance was comparable on test sets drawn from each dataset, the ensemble trained on the larger dataset retained its accuracy when these sets were crossed over, and when applied to a further collection of approximately 12,000 previously unseen structures introduced after all development work had been completed. In contrast, the ensemble trained on the smaller dataset showed a notable decline in generalisation ability. This difference is attributed to the greater diversity of atomic environments captured in the larger dataset. The larger dataset also enabled more robust modelling of various error properties, providing a quantitative foundation for spectral assignment and verification. This was achieved in two ways. First, a clear relationship was observed between prediction errors and the frequency of different node feature vectors in the training data, allowing error estimates to be associated with individual nodes based on their type. These estimates can be used as weights in a modified cityblock distance metric when assigning observed to predicted shifts. Second, the mean absolute prediction error calculated at the structure level is well-fitted by a Gaussian kernel cumulative distribution. This enabled a probabilistic assessment of whether the predicted shifts and assigned observations are consistent with originating from the same molecular structure.

Conformational Sampling in Computational Studies of Natural Products: Why Is It Important?

Conformational sampling is a vital component of a reliable computational chemistry investigation. With the aim of illustrating the importance of conformational sampling, and building awareness among new practitioners, we present a series of case studies that show how the quality and reliability of computational studies depend on undertaking a thorough conformer search. The examples are drawn from the most common types of research questions in natural products chemistry, but the fundamental principles apply more generally to computational studies of molecular structure and behavior in any field of chemistry.

Deciphering molecular structures: NMR spectroscopy and quantum mechanical insights of halogenated 4H-Chromenediones

Sesquiterpene lactones (SL) represent a class of secondary metabolites found in the Asteraceae family, notable for their unique structures. The SL α-santonin (1) and its derivatives are worthy of mention due to their diverse biological properties. Additionally, 4H-chromenes and 4H-chromones are appealing frameworks holding the capability to be used as structural motifs for new drugs. Furthermore, unambiguous structural elucidation is crucial for developing novel compounds for diverse applications. In this context, it is common to find in the literature molecules erroneously assigned. Therefore, the use of quantum mechanical calculations to simulate NMR chemical shifts has emerged as a valuable strategy. In this work, we conceived the synthesis of two halogenated 4H-chromenediones derived from photosantonic acid (2), a photoproduct arising from irradiation of α-santonin (1) in the ultraviolet region. The structure of the chlorinated and brominated products was determined by NMR analysis, with the aid of quantum mechanical calculations at the B3LYP/6-311 + G(2d,p)//M062x/6-31 + G(d,p) level of theory. All analyses were in agreement and led to the assignment of the brominated 4H-chromene-2,7-dione as (3S,3aS,5aR,9bS)-5a-(2-bromopropan-2-yl)-3-methyl-3,3a,5,5a,8,9b-hexahydro-4H-furo[2,3-f]chromene-2,7-dione (11b) and of the chlorinated 4H-chromene-2,7-dione as (3S,3aS,5aR,9bS)-5a-(2-chloropropan-2-yl)-3-methyl-3,3a,5,5a,8,9b-hexahydro-4H-furo[2,3-f]chromene-2,7-dione (12b). The diastereoselectivities of the reactions were explained based on products and intermediates formation energy calculated using B3LYP/6-31 + G(d,p) as the level of theory. Structures 11b and 12b were identified as the thermodynamic and kinetic products of the reaction among all candidates. Consequently, the strategy utilized in this study is robust and successfully illustrates the use of quantum mechanical calculations in the structural elucidation of new compounds with potential applications as novel drugs or products.

Reassignment of the Structure of a Tryptophan-Containing Cyclic Tripeptide Produced by the Biarylitide Crosslinking Cytochrome P450blt

The structure of the sidechain crosslinked Tyr-Leu-Trp peptide produced by the biarylitide crosslinking cytochrome P450Blt from Micromonospora sp. MW-13 has been reanalysed by a series of NMR, computational and isotope labelling experiments and shown to contain a C-N rather than a C-O bond. Additional in vivo experiments using such a modified peptide show there is a general tolerance of biarylitide crosslinking P450 enzymes for histidine to tryptophan mutations within their minimal peptide substrate sequences despite the lack of such residues noted in natural biarylitide gene clusters. This work further highlights the impressive ability of P450s from biarylitide biosynthesis pathways to act as biocatalysts for the formation of a range of sidechain crosslinked tripeptides.

An Integrated ANN-PRA/DP4+ Tandem Computational Approach Contributing to the Ordering of the Heliannuol Family

Heliannuols are a unique class of sesquiterpenes isolated mostly from Helianthus annuus, commonly known as sunflower. The interesting allelopathic properties, combined with their unprecedented carbon skeletons, have drawn wide attention to phytochemistry and synthetic groups. So far, 14 heliannuols (heliannuols A-N) have been described in the literature, although some of them have not yet been validated by total synthesis. Moreover, the structural proposal of some compounds was based on the similarity of NMR data reported for previously isolated analogues (which in many instances turned out to be incorrect), coupled with little or no stereochemical analysis. Consequently, the structural reassignment is a recurring theme in heliannuol's family. Through a rigorous and comprehensive quantum chemical simulation of NMR parameters, encompassing an integrated ANN-PRA/DP4+ tandem approach, we intended to advance unexplored directions regarding the structure of the entire heliannuol family. Furthermore, we found that the size of the fused ring significantly influences the signals corresponding to the aromatic ring, making this discovery an excellent diagnostic tool for quickly determining the core structure of these compounds.

A Very Deep Graph Convolutional Network for 13C NMR Chemical Shift Calculations with Density Functional Theory Level Performance for Structure Assignment

Nuclear magnetic resonance (NMR) chemical shift calculations are powerful tools for structure elucidation and have been extensively employed in both natural product and synthetic chemistry. However, density functional theory (DFT) NMR chemical shift calculations are usually time-consuming, while fast data-driven methods often lack reliability, making it challenging to apply them to computationally intensive tasks with a high requirement on quality. Herein, we have constructed a 54-layer-deep graph convolutional network for 13C NMR chemical shift calculations, which achieved high accuracy with low time-cost and performed competitively with DFT NMR chemical shift calculations on structure assignment benchmarks. Our model utilizes a semiempirical method, GFN2-xTB, and is compatible with a broad variety of organic systems, including those composed of hundreds of atoms or elements ranging from H to Rn. We used this model to resolve the controversial J/K ring junction problem of maitotoxin, which is the largest whole molecule assigned by NMR calculations to date. This model has been developed into user-friendly software, providing a useful tool for routine rapid structure validation and assignation as well as a new approach to elucidate the large structures that were previously unsuitable for NMR calculations.

HSQC Spectra Simulation and Matching for Molecular Identification

In the pursuit of improved compound identification and database search tasks, this study explores heteronuclear single quantum coherence (HSQC) spectra simulation and matching methodologies. HSQC spectra serve as unique molecular fingerprints, enabling a valuable balance of data collection time and information richness. We conducted a comprehensive evaluation of the following four HSQC simulation techniques: ACD/Labs (ACD), MestReNova (MNova), Gaussian NMR calculations (DFT), and a graph-based neural network (ML). For the latter two techniques, we developed a reconstruction logic to combine proton and carbon 1D spectra into HSQC spectra. The methodology involved the implementation of three peak-matching strategies (minimum-sum, Euclidean-distance, and Hungarian distance) combined with three padding strategies (zero-padding, peak-truncated, and nearest-neighbor double assignment). We found that coupling these strategies with a robust simulation technique facilitates the accurate identification of correct molecules from similar analogues (regio- and stereoisomers) and allows for fast and accurate large database searches. Furthermore, we demonstrated the efficacy of the best-performing methodology by rectifying the structures of a set of previously misidentified molecules. This research indicates that effective HSQC spectral simulation and matching methodologies significantly facilitate molecular structure elucidation. Furthermore, we offer a Google Colab notebook for researchers to use our methods on their own data (https://github.com/AstraZeneca/hsqc_structure_elucidation.git).

Study of heavy atom influence on poly-halogenated compounds using DP4/MM-DP4+/DP4+: insights and trends

Nuclear magnetic resonance (NMR) spectroscopy complemented by density functional theory (DFT) calculations is a crucial tool for structural elucidation. Nevertheless, the precision of NMR predictions is influenced by the 'heavy atom effect', wherein heavy atoms affect the shielding values of neighboring light atoms (HALA effect). Standard practice in the field involves removing the conflicting signals. However, in the case of polyhalogenated molecules, this is challenging due to the significant amount of information that ends up being lost. In this study the HALA is thoroughly investigated in the context of three leading probability methods: DP4, MM-DP4+, and DP4+. The results show that DP4+ is more sensitive to C-Cl or C-Br signals, which is a consequence of the longer bond lengths computed with DFT. Removing conflicting signals is highly effective in DP4+, but has an uncertain outcome in methods based on molecular mechanics geometries, such as DP4 and MM-DP4+. A detailed investigation of the effect of bond distance on the corresponding chemical shifts has also been conducted.

Road Map Toward Computer-Guided Total Synthesis of Natural Products. The Dysiherbol A Case Study: What if Serendipity Hadn't Intervened?

We present a computational study inspired by the story of dysiherbol A, a natural product whose putative structure was found incorrect through synthesis by a completely fortuitous event. While the carbon connectivity and chemical environment between both structures remain similar, the real dysiherbol A has a different molecular weight than that reported for the natural product. Had the synthesis groups not been favored by fortune, it could be speculated that a substantial amount of time and effort would have been required to solve the structural puzzle. Within the realm of computer-guided total synthesis of natural products, the question arose whether a synthesis group could have in silico reassigned the structure before embarking on the experimental adventure. To address this query, we evaluated some state-of-the-art computational procedures based on their computational demand and ease of implementation for nonexpert users with basic skills in computational chemistry (including HOSE, CASCADE, ANN-PRA, ML-J-DP4, DP4, and DP4+). While discussing the strengths and limitations of these methods, this case study provides a roadmap of what could be done before venturing into complex and time-demanding total synthesis projects.

Simple, accurate, adjustable-parameter-free prediction of NMR shifts for molecules in solution

Accurate prediction of NMR shifts is invaluable for interpreting and assigning NMR spectra, especially for complex applications such as determining the identity of unknown substances or resolving stereochemical assignments. Statistical linear regression models have proven effective for accurately correlating density functional theory predictions of chemical shieldings with experimentally-measured shifts, but lack transferability - they must be reparameterised using a reasonably extensive training set at each level of theory and for each choice of NMR solvent. We have previously introduced a novel two-point "shift-and-scale" correction procedure for gas phase shieldings that overcomes these limitations without significant loss of accuracy. In this work, we demonstrate that this approach is equally applicable for predicting solution-phase shifts from computed gas phase shieldings, using acetaldehyde as an experimentally and computationally convenient reference system. We also present all of the required experimental reference data to enable this approach to be used for any target analyte in a range of commonly used NMR solvents (chloroform, dichloromethane, acetonitrile, methanol, acetone, DMSO, D2O, benzene, pyridine).

Epimeric Mixture Analysis and Absolute Configuration Determination Using an Integrated Spectroscopic and Computational Approach-A Case Study of Two Epimers of 6-Hydroxyhippeastidine

Structural elucidation has always been challenging, and misassignment remains a stringent issue in the field of natural products. The growing interest in discovering unknown, complex natural structures accompanies the increasing awareness concerning misassignments in the community. The combination of various spectroscopic methods with molecular modeling has gained popularity in recent years. In this work, we demonstrated, for the first time, its power to fully elucidate the 2-dimensional and 3-dimensional structures of two epimers in an epimeric mixture of 6-hydroxyhippeastidine. DFT calculation of chemical shifts was first performed to assist the assignment of planar structures. Furthermore, relative and absolute configurations were established by three different ways of computer-assisted structure elucidation (CASE) coupled with ORD/ECD/VCD spectroscopies. In addition, the significant added value of OR/ORD computations to relative and absolute configuration determination was also revealed. Remarkably, the differentiation of two enantiomeric scaffolds (crinine and haemanthamine) was accomplished via OR/ORD calculations with cross-validation by ECD and VCD.

Lycojapomines A-E: Lycopodium Alkaloids with Anti-Renal Fibrosis Potential from Lycopodium japonicum

Five Lycopodium alkaloids featuring novel C₁₇N₂ (1 and 2), C₂₉N₃ (3 and 4), and C₁₅N₂ (5) skeletons were isolated from Lycopodium japonicum. Compound 1 is the first natural product containing a 3-aza[3.3.3]propellane motif. The structures of these compounds were elucidated by spectroscopic analysis, X-ray crystallography, and computational methods. Compounds 1 and 3-5 significantly inhibited TGF-β₁-induced fibronectin deposition in HK-2 cells at a nontoxic concentration of 20 μM.