General Protocol for the Accurate Prediction of Molecular 13C/1H NMR Chemical Shifts via Machine Learning Augmented DFT

A machine learning approach for dynamical modelling of Al distributions in zeolites via23Na/27Al solid-state NMR

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 was found to lead to errors of several ppm, and has 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 the 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 of dynamics on the spectra, which is often overlooked in computational studies despite its clear manifestation in experimental setups.

Review of machine learning for lipid nanoparticle formulation and process development

Lipid nanoparticles (LNPs) are a subset of pharmaceutical nanoparticulate formulations designed to encapsulate, stabilize, and deliver nucleic acid cargoes in vivo. Applications for LNPs include new interventions for genetic disorders, novel classes of vaccines, and alternate modes of intracellular delivery for therapeutic proteins. In the pharmaceutical industry, establishing a robust formulation and process to achieve target product performance is a critical component of drug development. Fundamental understanding of the processes for making LNPs and their interactions with biological systems have advanced considerably in the wake of the COVID-19 pandemic. Nevertheless, LNP formulation research remains largely empirical and resource intensive due to the multitude of input parameters and the complex physical phenomena that govern the processes of nanoparticle precipitation, self-assembly, structure evolution, and stability. Increasingly, artificial intelligence and machine learning (AI/ML) are being applied to improve the efficiency of research activities through in silico models and predictions, and to drive deeper fundamental understanding of experimental inputs to functional outputs. This review will identify current challenges and opportunities in the development of robust LNP formulations of nucleic acids, review studies that apply machine learning methods to experimental datasets, and provide discussion on associated data science challenges to facilitate collaboration between formulation and data scientists, aiming to accelerate the advancement of AI/ML applied to LNP formulation and process optimization.

Bent naphthodithiophenes: synthesis and characterization of isomeric fluorophores

Thiophene-containing heteroarenes are one of the most well-known classes of π-conjugated building blocks for photoactive molecules. Isomeric naphthodithiophenes (NDTs) are at the forefront of this research area due to their straightforward synthesis and derivatization. Notably, NDT geometries that are bent - such as naphtho[2,1-b:3,4-b']dithiophene (α-NDT) and naphtho[1,2-b:4,3-b']dithiophene (β-NDT) - are seldom employed as photoactive small molecules. This report investigates how remote substituents impact the photophysical properties of isomeric α- and β-NDTs. The orientation of the thiophene units plays a critical role in the emission: in the α(OHex)R2 series conjugation from the end-caps to the NDT core is apparent, while in the β(Oi-Pent)R2 series minimal change is observed unless strong electron acceptors, such as β(Oi-Pent)(PhCF3)2, are employed. This push-pull acceptor-donor-acceptor (A-D-A) fluorophore exhibits positive fluorosolvatochromism that correlates with increasing solvent polarity parameter, E T(30). In total, these results highlight how remote substituents are able to modulate the emission of isomeric bent NDTs.

Introducing 'identification probability' for automated and transferable assessment of metabolite identification confidence in metabolomics and related studies

Methods for assessing compound identification confidence in metabolomics and related studies have been debated and actively researched for the past two decades. The earliest effort in 2007 focused primarily on mass spectrometry and nuclear magnetic resonance spectroscopy and resulted in four recommended levels of metabolite identification confidence - the Metabolite Standards Initiative (MSI) Levels. In 2014, the original MSI Levels were expanded to five levels (including two sublevels) to facilitate communication of compound identification confidence in high resolution mass spectrometry studies. Further refinement in identification levels have occurred, for example to accommodate use of ion mobility spectrometry in metabolomics workflows, and alternate approaches to communicate compound identification confidence also have been developed based on identification points schema. However, neither qualitative levels of identification confidence nor quantitative scoring systems address the degree of ambiguity in compound identifications in context of the chemical space being considered, are easily automated, or are transferable between analytical platforms. In this perspective, we propose that the metabolomics and related communities consider identification probability as an approach for automated and transferable assessment of compound identification and ambiguity in metabolomics and related studies. Identification probability is defined simply as 1/N, where N is the number of compounds in a reference library or chemical space that match to an experimentally measured molecule within user-defined measurement precision(s), for example mass measurement or retention time accuracy, etc. We demonstrate the utility of identification probability in an in silico analysis of multi-property reference libraries constructed from the Human Metabolome Database and computational property predictions, provide guidance to the community in transparent implementation of the concept, and invite the community to further evaluate this concept in parallel with their current preferred methods for assessing metabolite identification confidence.

Construction of Highly Accurate Machine Learning Potential Energy Surfaces for Excited-State Dynamics Simulations Based on Low-Level Data Sets

Machine learning is capable of effectively predicting the potential energies of molecules in the presence of high-quality data sets. Its application in the construction of ground- and excited-state potential energy surfaces is attractive to accelerate nonadiabatic molecular dynamics simulations of photochemical reactions. Because of the huge computational cost of excited-state electronic structure calculations, the construction of a high-quality data set becomes a bottleneck. In the present work, we first built two data sets. One was obtained from surface hopping dynamics simulations at the semiempirical OM2/MRCI level. Another was extracted from the dynamics trajectories at the CASSCF level, which was reported previously. The ground- and excited-state potential energy surfaces of ethylene-bridged azobenzene at the CASSCF computational level were constructed based on the former low-level data set. Although non-neural network machine learning methods can achieve good or modest performance during the training process, only neural network models provide reliable predictions on the latter external test data set. The BPNN and SchNet combined with the Δ-ML scheme and the force term in the loss functions are recommended for dynamics simulations. Then, we performed excited-state dynamics simulations of the photoisomerization of ethylene-bridged azobenzene on machine learning potential energy surfaces. Compared with the lifetimes of the first excited state (S1) estimated at different computational levels, our results on the E isomer are in good agreement with the high-level estimation. However, the overestimation of the Z isomer is unimproved. It suggests that smaller errors during the training process do not necessarily translate to more accurate predictions on high-level potential energies or better performance on nonadiabatic dynamics simulations, at least in the present case.

Can Graph Machines Accurately Estimate 13C NMR Chemical Shifts of Benzenic Compounds?

In the organic laboratory, the 13C nuclear magnetic resonance (NMR) spectrum of a newly synthesized compound remains an essential step in elucidating its structure. For the chemist, the interpretation of such a spectrum, which is a set of chemical-shift values, is made easier if he/she has a tool capable of predicting with sufficient accuracy the carbon-shift values from the structure he/she intends to prepare. As there are few open-source methods for accurately estimating this property, we applied our graph-machine approach to build models capable of predicting the chemical shifts of carbons. For this study, we focused on benzene compounds, building an optimized model derived from training a database of 10,577 chemical shifts originating from 2026 structures that contain up to ten types of non-carbon atoms, namely H, O, N, S, P, Si, and halogens. It provides a training root-mean-squared relative error (RMSRE) of 0.5%, i.e., a root-mean-squared error (RMSE) of 0.6 ppm, and a mean absolute error (MAE) of 0.4 ppm for estimating the chemical shifts of the 10k carbons. The predictive capability of the graph-machine model is also compared with that of three commercial packages on a dataset of 171 original benzenic structures (1012 chemical shifts). The graph-machine model proves to be very efficient in predicting chemical shifts, with an RMSE of 0.9 ppm, and compares favorably with the RMSEs of 3.4, 1.8, and 1.9 ppm computed with the ChemDraw v. 23.1.1.3, ACD v. 11.01, and MestReNova v. 15.0.1-35756 packages respectively. Finally, a Docker-based tool is proposed to predict the carbon chemical shifts of benzenic compounds solely from their SMILES codes.

Accurate Prediction of NMR Chemical Shifts: Integrating DFT Calculations with Three-Dimensional Graph Neural Networks

Computer prediction of NMR chemical shifts plays an increasingly important role in molecular structure assignment and elucidation for organic molecule studies. Density functional theory (DFT) and gauge-including atomic orbital (GIAO) have established a framework to predict NMR chemical shifts but often at a significant computational expense with a limited prediction accuracy. Recent advancements in deep learning methods, especially graph neural networks (GNNs), have shown promise in improving the accuracy of predicting experimental chemical shifts, either by using 2D molecular topological features or 3D conformational representation. This study presents a new 3D GNN model to predict 1H and 13C chemical shifts, CSTShift, that combines atomic features with DFT-calculated shielding tensor descriptors, capturing both isotropic and anisotropic shielding effects. Utilizing the NMRShiftDB2 data set and conducting DFT optimization and GIAO calculations at the B3LYP/6-31G(d) level, we prepared the NMRShiftDB2-DFT data set of high-quality 3D structures and shielding tensors with corresponding experimentally measured 1H and 13C chemical shifts. The developed CSTShift models achieve the state-of-the-art prediction performance on both the NMRShiftDB2-DFT test data set and external CHESHIRE data set. Further case studies on identifying correct structures from two groups of constitutional isomers show its capability for structure assignment and elucidation. The source code and data are accessible at https://yzhang.hpc.nyu.edu/IMA.

From NMR to AI: Designing a Novel Chemical Representation to Enhance Machine Learning Predictions of Physicochemical Properties

A novel approach to the utilization of nuclear magnetic resonance (NMR) spectroscopy data in the prediction of logD through machine learning algorithms is shown. In the analysis, a data set of 754 chemical compounds, organized into 30 clusters, was evaluated using advanced machine learning models, such as Support Vector Regression (SVR), Gradient Boosting, and AdaBoost, and comprehensive validation and testing methods were employed, including 10-fold cross-validation, bootstrapping, and leave-one-out. The study revealed the superior performance of the Bucket Integration method for dimensionality reduction, consistently yielding the lowest root mean square error (RMSE) across all data sets and normalization schemes. The SVR prediction models demonstrated remarkable computational efficiency and low cost, with the best RMSE value reaching 0.66. Our best model outperformed existing tools like JChem Suite's logD Predictor (0.91) and CplogD (1.27), and a comparison with traditional molecular representations yielded a comparable RMSE (0.50), emphasizing the robustness of our NMR data integration. The widespread availability of NMR data in pharmaceutical and industrial research presents an untapped resource for predictive modeling, highlighting the need for accessible methodologies like ours that complement the analytical toolbox beyond conventional 2D approaches. Our approach, designed to leverage the rich spatial data from NMR spectroscopy, provides additional insights and enriches drug discovery and computational chemistry with a freely accessible tool.

Perspectives on Nuclear Magnetic Resonance Spectroscopy in Drug Discovery Research

The drug discovery landscape has undergone a significant transformation over the past decade, owing to research endeavors in a wide range of areas leading to strategies for pursuing new drug targets and the emergence of novel drug modalities. NMR spectroscopy has been a technology of fundamental importance to these research pursuits and has seen its use expanded both within and outside of traditional medicinal chemistry applications. In this perspective, we will present advancement of NMR-derived methods that have facilitated the characterization of small molecules and novel drug modalities including macrocyclic peptides, cyclic dinucleotides, and ligands for protein degradation. We will discuss innovations in NMR spectroscopy at the chemistry and biology interface that have broadened NMR's utility from hit identification through lead optimization activities. We will also discuss the promise of emerging NMR approaches in bridging our understanding and addressing challenges in the pursuit of the therapeutic agents of the future.

Machine learning-based correction for spin-orbit coupling effects in NMR chemical shift calculations

As one of the most powerful analytical methods for molecular and solid-state structure elucidation, NMR spectroscopy is an integral part of chemical laboratories associated with a great research interest in its computational simulation. Particularly when heavy atoms are present, a relativistic treatment is essential in the calculations as these influence also the nearby light atoms. In this work, we present a Δ-machine learning method that approximates the contribution to 13C and 1H NMR chemical shifts that stems from spin-orbit (SO) coupling effects. It is built on computed reference data at the spin-orbit zeroth-order regular approximation (ZORA) DFT level for a set of 6388 structures with 38 740 13C and 64 436 1H NMR chemical shifts. The scope of the methods covers the 17 most important heavy p-block elements that exhibit heavy atom on the light atom (HALA) effects to covalently bound carbon or hydrogen atoms. Evaluated on the test data set, the approach is able to recover roughly 85% of the SO contribution for 13C and 70% for 1H from a scalar-relativistic PBE0/ZORA-def2-TZVP calculation at virtually no extra computational costs. Moreover, the method is transferable to other baseline DFT methods even without retraining the model and performs well for realistic organotin and -lead compounds. Finally, we show that using a combination of the new approach with our previous Δ-ML method for correlation contributions to NMR chemical shifts, the mean absolute NMR shift deviations from non-relativistic DFT calculations to experimental values can be halved.

QM assisted ML for 19F NMR chemical shift prediction

BACKGROUND

Ligand-observed 19F NMR detection is an efficient method for screening libraries of fluorinated molecules in fragment-based drug design campaigns. Screening fluorinated molecules in large mixtures makes 19F NMR a high-throughput method. Typically, these mixtures are generated from pools of well-characterized fragments. By predicting 19F NMR chemical shift, mixtures could be generated for arbitrary fluorinated molecules facilitating for example focused screens.

METHODS

In a previous publication, we introduced a method to predict 19F NMR chemical shift using rooted fluorine fingerprints and machine learning (ML) methods. Having observed that the quality of the prediction depends on similarity to the training set, we here propose to assist the prediction with quantum mechanics (QM) based methods in cases where compounds are not well covered by a training set.

RESULTS

Beyond similarity, the performance of ML methods could be associated with individual features in compounds. A combination of both could be used as a procedure to split input data sets into those that could be predicted by ML and those that required QM processing. We could show on a proprietary fluorinated fragment library, known as LEF (Local Environment of Fluorine), and a public Enamine data set of 19F NMR chemical shifts that ML and QM methods could synergize to outperform either method individually. Models built on Enamine data, as well as model building and QM workflow tools, can be found at https://github.com/PatrickPenner/lefshift and https://github.com/PatrickPenner/lefqm .

Frontiers of molecular crystal structure prediction for pharmaceuticals and functional organic materials

The reliability of organic molecular crystal structure prediction has improved tremendously in recent years. Crystal structure predictions for small, mostly rigid molecules are quickly becoming routine. Structure predictions for larger, highly flexible molecules are more challenging, but their crystal structures can also now be predicted with increasing rates of success. These advances are ushering in a new era where crystal structure prediction drives the experimental discovery of new solid forms. After briefly discussing the computational methods that enable successful crystal structure prediction, this perspective presents case studies from the literature that demonstrate how state-of-the-art crystal structure prediction can transform how scientists approach problems involving the organic solid state. Applications to pharmaceuticals, porous organic materials, photomechanical crystals, organic semi-conductors, and nuclear magnetic resonance crystallography are included. Finally, efforts to improve our understanding of which predicted crystal structures can actually be produced experimentally and other outstanding challenges are discussed.

Molecular Graph-Based Deep Learning Algorithm Facilitates an Imaging-Based Strategy for Rapid Discovery of Small Molecules Modulating Biomolecular Condensates

Biomolecular condensates are proposed to cause diseases, such as cancer and neurodegeneration, by concentrating proteins at abnormal subcellular loci. Imaging-based compound screens have been used to identify small molecules that reverse or promote biomolecular condensates. However, limitations of conventional imaging-based methods restrict the screening scale. Here, we used a graph convolutional network (GCN)-based computational approach and identified small molecule candidates that reduce the nuclear liquid-liquid phase separation of TAR DNA-binding protein 43 (TDP-43), an essential protein that undergoes phase transition in neurodegenerative diseases. We demonstrated that the GCN-based deep learning algorithm is suitable for spatial information extraction from the molecular graph. Thus, this is a promising method to identify small molecule candidates with novel scaffolds. Furthermore, we validated that these candidates do not affect the normal splicing function of TDP-43. Taken together, a combination of an imaging-based screen and a GCN-based deep learning method dramatically improves the speed and accuracy of the compound screen for biomolecular condensates.

Machine learning-assisted structure annotation of natural products based on MS and NMR data

Covering: up to March 2023Machine learning (ML) has emerged as a popular tool for analyzing the structures of natural products (NPs). This review presents a summary of the recent advancements in ML-assisted mass spectrometry (MS) and nuclear magnetic resonance (NMR) data analysis to establish the chemical structures of NPs. First, ML-based MS/MS analyses that rely on library matching are discussed, which involves the utilization of ML algorithms to calculate similarity, predict the MS/MS fragments, and form molecular fingerprint. Then, ML assisted MS/MS structural annotation without library matching is reviewed. Furthermore, the cases of ML algorithms in assisting structural studies of NPs based on NMR are discussed from four perspectives: NMR prediction, functional group identification, structural categorization and quantum chemical calculation. Finally, the review concludes with a discussion of the challenges and the trends associated with the structural establishment of NPs based on ML algorithms.

Advances in the Application of Artificial Intelligence-Based Spectral Data Interpretation: A Perspective

The interpretation of spectral data, including mass, nuclear magnetic resonance, infrared, and ultraviolet-visible spectra, is critical for obtaining molecular structural information. The development of advanced sensing technology has multiplied the amount of available spectral data. Chemical experts must use basic principles corresponding to the spectral information generated by molecular fragments and functional groups. This is a time-consuming process that requires a solid professional knowledge base. In recent years, the rapid development of computer science and its applications in cheminformatics and the emergence of computer-aided expert systems have greatly reduced the difficulty in analyzing large quantities of data. For expert systems, however, the problem-solving strategy must be known in advance or extracted by human experts and translated into algorithms. Gratifyingly, the development of artificial intelligence (AI) methods has shown great promise for solving such problems. Traditional algorithms, including the latest neural network algorithms, have shown great potential for both extracting useful information and processing massive quantities of data. This Perspective highlights recent innovations covering all of the emerging AI-based spectral interpretation techniques. In addition, the main limitations and current obstacles are presented, and the corresponding directions for further research are proposed. Moreover, this Perspective gives the authors' personal outlook on the development and future applications of spectral interpretation.

Preparative three-dimensional GC and nuclear magnetic resonance for the isolation and identification of two sesquiterpene ethers from Dictyota Dichotoma

Separation science plays a crucial role in the isolation of novel compounds contained in complex matrices. Yet their rationale employment needs preliminary structure elucidation, which usually requires sufficient aliquots of grade substances to characterize the molecule by nuclear magnetic resonance experiments. In this study, two peculiar oxa-tricycloundecane ethers were isolated by means of preparative multidimensional gas chromatography from the brown alga species Dictyota dichotoma (Huds.) Lam., aiming to assign their 3D structures. Density functional theory simulations were carried out to select the correct configurational species matching the experimental NMR data (in terms of enantiomeric couples). In this case, the theoretical approach was crucial as the protonic signal overlap and spectral overcrowding were preventing any other unambiguous structural information. Just after the identification through the density functional theory data matching of the correct relative configuration it was possible to verify an enhanced self-consistency with the experimental data, confirming the stereochemistry. The results obtained further pave the way toward structure elucidation of highly asymmetric molecules, whose configuration cannot be inferred by other means or strategies.

Atomic-level structure determination of amorphous molecular solids by NMR

Structure determination of amorphous materials remains challenging, owing to the disorder inherent to these materials. Nuclear magnetic resonance (NMR) powder crystallography is a powerful method to determine the structure of molecular solids, but disorder leads to a high degree of overlap between measured signals, and prevents the unambiguous identification of a single modeled periodic structure as representative of the whole material. Here, we determine the atomic-level ensemble structure of the amorphous form of the drug AZD4625 by combining solid-state NMR experiments with molecular dynamics (MD) simulations and machine-learned chemical shifts. By considering the combined shifts of all 1H and 13C atomic sites in the molecule, we determine the structure of the amorphous form by identifying an ensemble of local molecular environments that are in agreement with experiment. We then extract and analyze preferred conformations and intermolecular interactions in the amorphous sample in terms of the stabilization of the amorphous form of the drug.

Total Synthesis and Absolute Configuration Determination of the α-Glycosidase Inhibitor (3S,4R)-6-Acetyl-3-hydroxy-2,2-dimethylchroman-4-yl (Z)-2-Methylbut-2-enoate from Ageratina grandifolia

Herein, we report the first total synthesis of α-glycosidase inhibitor (3R, 4S)-6-acetyl-3-hydroxy-2,2-dimethylchroman-4-yl (Z)-2-methylbut-2-enoate as well as its enantiomer. Our synthesis confirms the chromane structure separately proposed by Navarro-Vazquez and Mata, on the basis of DFT computations. Furthermore, our synthesis allowed us to determine the absolute configuration of the natural compound as (3S, 4R) and not (3R, 4S).

Computation of CCSD(T)-Quality NMR Chemical Shifts via Δ-Machine Learning from DFT

NMR spectroscopy undoubtedly plays a central role in determining molecular structures across different chemical disciplines, and the accurate computational prediction of NMR parameters is highly desirable. In this work, a new Δ-machine learning approach is presented to correct DFT-computed NMR chemical shifts using input features from the calculation and in addition highly accurate reference data at the CCSD(T)/pcSseg-2 level of theory with a basis set extrapolation scheme. The model is trained on a data set containing 1000 optimized and geometrically distorted structures of small organic molecules comprising most elements of the first three periods and containing data for 7090 ¹H and 4230 ¹³C NMR chemical shifts. Applied to the PBE0/pcSseg-2 method, the mean absolute deviation (MAD) on the internal NMR shift test set is reduced by 81% for ¹H and 92% for ¹³C at virtually no additional computational cost. For 12 different DFT functional and basis set combinations, the MAD of the ML-corrected NMR shifts ranges from 0.021 to 0.039 ppm (¹H) and from 0.38 to 1.07 ppm (¹³C). Importantly, the new method consistently outperforms the simple and widely used linear regression correction technique. This behavior is reproduced on three different external benchmark sets, confirming the generality and robustness of the correction scheme, which can easily be applied in DFT-based spectral simulations.

Do Models beyond Hybrid Density Functionals Increase the Agreement with Experiment for Predicted NMR Chemical Shifts or Electric Field Gradient Tensors in Organic Solids?

Ab initio predictions of chemical shifts and electric field gradient (EFG) tensor components are frequently used to help interpret solid-state nuclear magnetic resonance (NMR) experiments. Typically, these predictions employ density functional theory (DFT) with generalized gradient approximation (GGA) functionals, though hybrid functionals have been shown to improve accuracy relative to experiment. Here, the performance of a dozen models beyond the GGA approximation are examined for the prediction of solid-state NMR observables, including meta-GGA, hybrid, and double-hybrid density functionals and second-order Møller-Plesset perturbation theory (MP2). These models are tested on organic molecular crystal data sets containing 169 experimental ¹³C and ¹⁵N chemical shifts and 114 ¹⁷O and ¹⁴N EFG tensor components. To make these calculations affordable, gauge-including projector augmented wave (GIPAW) Perdew-Burke-Ernzerhof (PBE) calculations with periodic boundary conditions are combined with a local intramolecular correction computed at the higher level of theory. Within the context of typical NMR property calculations performed on a static, DFT-optimized crystal structure, the benchmarking finds that the double-hybrid DFT functionals produce errors versus experiment that are no smaller than those of hybrid functionals in the best cases, and they can be larger. MP2 errors versus experiment are even bigger. Overall, no practical advantages are found for using any of the tested double-hybrid functionals or MP2 to predict experimental solid-state NMR chemical shifts and EFG tensor components for routine organic crystals, especially given the higher computational cost of those methods. This finding likely reflects error cancellation benefiting the hybrid functionals. Improving the accuracy of the predicted chemical shifts and EFG tensors relative to experiment would probably require more robust treatments of the crystal structures, their dynamics, and other factors.

Accurate prediction of 11B NMR chemical shift of BODIPYs via machine learning

In this article, we present the results of developing a model based on an RFR machine learning method using the ISIDA fragment descriptors for predicting the ¹¹B NMR chemical shift of BODIPYs. The model is freely available at https://ochem.eu/article/146458. The model demonstrates the high quality of predicting the ¹¹B NMR chemical shift (RMSE, 5CV (FINALE training set) = 0.40 ppm, RMSE (TEST set) = 0.14 ppm). In addition, we compared the "cost" and the user-friendliness for calculations using the quantum-chemical model with the DFT/GIAO approach. The ¹¹B NMR chemical shift prediction accuracy (RMSE) of the model considered is more than three times higher and tremendously faster than the DFT/GIAO calculations. As a result, we provide a convenient tool and database that we collected for all researchers, that allows them to predict the ¹¹B NMR chemical shift of boron-containing dyes. We believe that the new model will make it easier for researchers to correctly interpret the ¹¹B NMR chemical shifts experimentally determined and to select more optimal conditions to perform an NMR experiment.

Predicting 195 Pt NMR Chemical Shifts in Water-Soluble Inorganic/Organometallic Complexes with a Fast and Simple Protocol Combining Semiempirical Modeling and Machine Learning

Water-soluble Pt complexes are the key components in medicinal chemistry and catalysis. The well-known cisplatin family of anticancer drugs and industrial hydrosylilation catalysts are two leading examples. On the molecular level, the activity mechanisms of such complexes mostly involve changes in the Pt coordination sphere. Using ¹⁹⁵ Pt NMR spectroscopy for operando monitoring would be a valuable tool for uncovering the activity mechanisms; however, reliable approaches for the rapid correlation of Pt complex structure with ¹⁹⁵ Pt chemical shifts are very challenging and not available for everyday research practice. While NMR shielding is a response property, molecular 3D structure determines NMR spectra, as widely known, which allows us to build up 3D structure to ¹⁹⁵ Pt chemical shift correlations. Accordingly, we present a new workflow for the determination of lowest-energy configurational/conformational isomers based on the GFN2-xTB semiempirical method and prediction of corresponding chemical shifts with a Machine Learning (ML) model tuned for Pt complexes. The workflow was designed for the prediction of ¹⁹⁵ Pt chemical shifts of water-soluble Pt(II) and Pt(IV) anionic, neutral, and cationic complexes with halide, NO₂ ^- , (di)amino, and (di)carboxylate ligands with chemical shift values ranging from -6293 to 7090 ppm. The model offered an accuracy (normalized root-mean-square deviation/RMSD) of 1.08 %/145.02 ppm on the held-out test set.

Primary Structure of Glycans by NMR Spectroscopy

Glycans, carbohydrate molecules in the realm of biology, are present as biomedically important glycoconjugates and a characteristic aspect is that their structures in many instances are branched. In determining the primary structure of a glycan, the sugar components including the absolute configuration and ring form, anomeric configuration, linkage(s), sequence, and substituents should be elucidated. Solution state NMR spectroscopy offers a unique opportunity to resolve all these aspects at atomic resolution. During the last two decades, advancement of both NMR experiments and spectrometer hardware have made it possible to unravel carbohydrate structure more efficiently. These developments applicable to glycans include, inter alia, NMR experiments that reduce spectral overlap, use selective excitations, record tilted projections of multidimensional spectra, acquire spectra by multiple receivers, utilize polarization by fast-pulsing techniques, concatenate pulse-sequence modules to acquire several spectra in a single measurement, acquire pure shift correlated spectra devoid of scalar couplings, employ stable isotope labeling to efficiently obtain homo- and/or heteronuclear correlations, as well as those that rely on dipolar cross-correlated interactions for sequential information. Refined computer programs for NMR spin simulation and chemical shift prediction aid the structural elucidation of glycans, which are notorious for their limited spectral dispersion. Hardware developments include cryogenically cold probes and dynamic nuclear polarization techniques, both resulting in enhanced sensitivity as well as ultrahigh field NMR spectrometers with a 1H NMR resonance frequency higher than 1 GHz, thus improving resolution of resonances. Taken together, the developments have made and will in the future make it possible to elucidate carbohydrate structure in great detail, thereby forming the basis for understanding of how glycans interact with other molecules.

Simulation Reveals the Chameleonic Behavior of Macrocycles

Conformational analysis is central to the design of bioactive molecules. It is particularly challenging for macrocycles due to noncovalent transannular interactions, steric interactions, and ring strain that are often coupled. Herein, we simulated the conformations of five macrocycles designed to express a progression of increasing complexity in environment-dependent intramolecular interactions and verified the results against NMR measurements in chloroform and dimethyl sulfoxide. Molecular dynamics using an explicit solvent model, but not the Monte Carlo method with implicit solvation, handled both solvents correctly. Refinement of conformations at the ab initio level was fundamental to reproducing the experimental observations─standard state-of-the-art molecular mechanics force fields were insufficient. Our simulations correctly predicted the intramolecular interactions between side chains and the macrocycle and revealed an unprecedented solvent-induced conformational switch of the macrocyclic ring. Our results provide a platform for the rational, prospective design of molecular chameleons that adapt to the properties of the environment.

Machine learning, artificial intelligence, and chemistry: how smart algorithms are reshaping simulation and the laboratory

Machine learning and artificial intelligence are increasingly gaining in prominence through image analysis, language processing, and automation, to name a few applications. Machine learning is also making profound changes in chemistry. From revisiting decades-old analytical techniques for the purpose of creating better calibration curves, to assisting and accelerating traditional in silico simulations, to automating entire scientific workflows, to being used as an approach to deduce underlying physics of unexplained chemical phenomena, machine learning and artificial intelligence are reshaping chemistry, accelerating scientific discovery, and yielding new insights. This review provides an overview of machine learning and artificial intelligence from a chemist’s perspective and focuses on a number of examples of the use of these approaches in computational chemistry and in the laboratory.

A Look Inside the Black Box of Machine Learning Photodynamics Simulations

ConspectusPhotochemical reactions are of great importance in chemistry, biology, and materials science because they take advantage of a renewable energy source, mild reaction conditions, and high atom economy. Light absorption can excite molecules to a higher energy electronic state of the same spin multiplicity. The following nonadiabatic processes induce molecular transformations that afford exotic molecular architectures and high-energy-isomers that are inaccessible by thermal means. Computational simulations now complement time-resolved instrumentation to reveal ultrafast excited-state mechanistic information for photochemical reactions that is essential in disentangling elusive spectroscopic features, excited-state lifetimes, and excited-state mechanistic critical points. Nonadiabatic molecular dynamics (NAMD), powered by surface hopping techniques, is among the most widely applied techniques to model the photochemical reactions of medium-sized molecules. However, the computational efficiency is limited because of the requisite thousands of multiconfigurational quantum-chemical calculations multiplied by hundreds of trajectories. Machine learning (ML) has emerged as a revolutionary force in computational chemistry to predict the outcome of the resource-intensive multiconfigurational calculations on the fly. An ML potential trained with a substantial set of quantum-chemical calculations can predict the energies and forces with errors under chemical accuracy at a negligible cost. The integration of ML potentials in NAMD dramatically extends the maximum simulation time scale by ∼10 000-fold to the nanosecond regime.In this Account, we present a comprehensive demonstration of ML photodynamics simulations and summarize our most recent applications in resolving complex photochemical reactions. First, we address three fundamental components of ML techniques for photodynamics simulations: the quantum-chemical data set, the ML potential, and NAMD. Second, we describe best practices in building training data and our procedure toward training the ML photodynamics model with our recent literature contributions. We introduce a convenient training data generation scheme combining Wigner sampling and geometrical interpolation. It trains reliable and effective ML potentials suitable for subsequent active learning to detect undersampled data. We demonstrate how active learning automatically discovers new mechanistic pathways and reproduces experimental results. We point out that atomic permutation is an essential data augmentation approach to improve the learnability of distance-based molecular descriptors for highly symmetric molecules. Third, we demonstrate the utility of ML-photodynamics by showing the results of ML photodynamics simulations of (1) photo-torquoselective 4π disrotatory electrocyclic ring closing of norbornyl cyclohexadiene, which reveals a thermal conversion from experimentally unobserved intermediates to the reactant in 1 ns; (2) [2 + 2] photocycloaddition of substituted [3]-syn-ladderdienes in competition with 4π and 6π electrocyclic ring-opening reactions, uncovering substituent effects to explain the reported increased quantum yield of substituted cubane precursors; and (3) photochemical 4π disrotatory electrocyclic reactions of fluorobenzenes in nanoseconds with XMS-CASPT2-level training data. We expect this Account to broaden understanding of ML photodynamics and inspire future developments and applications to increasingly large molecules within complex environments on long time scales.

Atom typing using graph representation learning: How do models learn chemistry?

Atom typing is the first step for simulating molecules using a force field. Automatic atom typing for an arbitrary molecule is often realized by rule-based algorithms, which have to manually encode rules for all types defined in this force field. These are time-consuming and force field-specific. In this study, a method that is independent of a specific force field based on graph representation learning is established for automatic atom typing. The topology adaptive graph convolution network (TAGCN) is found to be an optimal model. The model does not need manual enumeration of rules but can learn the rules just through training using typed molecules prepared during the development of a force field. The test on the CHARMM general force field gives a typing correctness of 91%. A systematic error of typing by TAGCN is its inability of distinguishing types in rings or acyclic chains. It originates from the fundamental structure of graph neural networks and can be fixed in a trivial way. More importantly, analysis of the rationalization processes of these models using layer-wise relation propagation reveals how TAGCN encodes rules learned during training. Our model is found to be able to type using the local chemical environments, in a way highly in accordance with chemists’ intuition.

Regression Machine Learning Models Used to Predict DFT-Computed NMR Parameters of Zeolites

Machine learning approaches can drastically decrease the computational time for the predictions of spectroscopic properties in materials, while preserving the quality of the computational approaches. We studied the performance of kernel-ridge regression (KRR) and gradient boosting regressor (GBR) models trained on the isotropic shielding values, computed with density-functional theory (DFT), in a series of different known zeolites containing out-of-frame metal cations or fluorine anion and organic structure-directing cations. The smooth overlap of atomic position descriptors were computed from the DFT-optimised Cartesian coordinates of each atoms in the zeolite crystal cells. The use of these descriptors as inputs in both machine learning regression methods led to the prediction of the DFT isotropic shielding values with mean errors within 0.6 ppm. The results showed that the GBR model scales better than the KRR model.

Graph Convolutional Network-Based Screening Strategy for Rapid Identification of SARS-CoV-2 Cell-Entry Inhibitors

The cell entry of SARS-CoV-2 has emerged as an attractive drug development target. We previously reported that the entry of SARS-CoV-2 depends on the cell surface heparan sulfate proteoglycan (HSPG) and the cortex actin, which can be targeted by therapeutic agents identified by conventional drug repurposing screens. However, this drug identification strategy requires laborious library screening, which is time consuming, and often limited number of compounds can be screened. As an alternative approach, we developed and trained a graph convolutional network (GCN)-based classification model using information extracted from experimentally identified HSPG and actin inhibitors. This method allowed us to virtually screen 170,000 compounds, resulting in ∼2000 potential hits. A hit confirmation assay with the uptake of a fluorescently labeled HSPG cargo further shortlisted 256 active compounds. Among them, 16 compounds had modest to strong inhibitory activities against the entry of SARS-CoV-2 pseudotyped particles into Vero E6 cells. These results establish a GCN-based virtual screen workflow for rapid identification of new small molecule inhibitors against validated drug targets.

DU8ML: Machine Learning-Augmented Density Functional Theory Nuclear Magnetic Resonance Computations for High-Throughput In Silico Solution Structure Validation and Revision of Complex Alkaloids

Machine learning (ML) profoundly improves the accuracy of the fast DU8+ hybrid density functional theory/parametric computations of nuclear magnetic resonance spectra, allowing for high throughput in silico validation and revision of complex alkaloids and other natural products. Of nearly 170 alkaloids surveyed, 35 structures are revised with the next-generation ML-augmented DU8 method, termed DU8ML.

An automated framework for high-throughput predictions of NMR chemical shifts within liquid solutions

Identifying stable speciation in multi-component liquid solutions is fundamentally important to areas from electrochemistry to organic chemistry and biomolecular systems. Here we introduce a fully automated, high-throughput computational framework for the accurate prediction of stable species in liquid solutions by computing the nuclear magnetic resonance (NMR) chemical shifts. The framework automatically extracts and categorizes hundreds of thousands of atomic clusters from classical molecular dynamics simulations, identifies the most stable species in solution and calculates their NMR chemical shifts via density functional theory calculations. Additionally, the framework creates a database of computed chemical shifts for liquid solutions across a wide chemical and parameter space. We compare our computational results to experimental measurements for magnesium bis(trifluoromethanesulfonyl)imide Mg(TFSI)2 salt in dimethoxyethane solvent. Our analysis of the Mg2+ solvation structural evolutions reveals key factors that influence the accuracy of NMR chemical shift predictions in liquid solutions. Furthermore, we show how the framework reduces the performance of over 300 13C and 600 1H density functional theory chemical shift predictions to a single submission procedure.

A general QSPR protocol for the prediction of atomic/inter-atomic properties: a fragment based graph convolutional neural network (F-GCN)

In this study, a general quantitative structure-property relationship (QSPR) protocol, fragment based graph convolutional neural network (F-GCN), was developed for the prediction of atomic/inter-atomic properties. We applied this novel artificial intelligence (AI) tool in predictions of NMR chemical shifts and bond dissociation energies (BDEs). The obtained results were comparable to experimental measurements, while the computational cost was substantially reduced, with respect to pure density functional theory (DFT) calculations. The two important features of F-GCN can be summarised as: first, it could utilise different levels of molecular fragments for atomic/inter-atomic information extraction; second, the designed architecture is also open to include additional descriptors for a more accurate solution of the local environment at atomic level, making itself more efficient for structural solutions. And during our test, the averaged prediction error of 1H NMR chemical shifts is as small as 0.32 ppm, and the error of C-H BDE estimation is 2.7 kcal mol-1. Moreover, we further demonstrated the applicability of this developed F-GCN model via several challenging structural assignments. The success of the F-GCN in atomic and inter-atomic predictions also indicates an essential improvement of computational chemistry with the assistance of AI tools.

Quantum Chemistry Calculations for Metabolomics

A primary goal of metabolomics studies is to fully characterize the small-molecule composition of complex biological and environmental samples. However, despite advances in analytical technologies over the past two decades, the majority of small molecules in complex samples are not readily identifiable due to the immense structural and chemical diversity present within the metabolome. Current gold-standard identification methods rely on reference libraries built using authentic chemical materials ("standards"), which are not available for most molecules. Computational quantum chemistry methods, which can be used to calculate chemical properties that are then measured by analytical platforms, offer an alternative route for building reference libraries, i.e., in silico libraries for "standards-free" identification. In this review, we cover the major roadblocks currently facing metabolomics and discuss applications where quantum chemistry calculations offer a solution. Several successful examples for nuclear magnetic resonance spectroscopy, ion mobility spectrometry, infrared spectroscopy, and mass spectrometry methods are reviewed. Finally, we consider current best practices, sources of error, and provide an outlook for quantum chemistry calculations in metabolomics studies. We expect this review will inspire researchers in the field of small-molecule identification to accelerate adoption of in silico methods for generation of reference libraries and to add quantum chemistry calculations as another tool at their disposal to characterize complex samples.

Revving up 13C NMR shielding predictions across chemical space: benchmarks for atoms-in-molecules kernel machine learning with new data for 134 kilo molecules

The requirement for accelerated and quantitatively accurate screening of nuclear magnetic resonance spectra across the small molecules chemical compound space is two-fold: (1) a robust ‘local’ machine learning (ML) strategy capturing the effect of the neighborhood on an atom’s ‘near-sighted’ property—chemical shielding; (2) an accurate reference dataset generated with a state-of-the-art first-principles method for training. Herein we report the QM9-NMR dataset comprising isotropic shielding of over 0.8 million C atoms in 134k molecules of the QM9 dataset in gas and five common solvent phases. Using these data for training, we present benchmark results for the prediction transferability of kernel-ridge regression models with popular local descriptors. Our best model, trained on 100k samples, accurately predicts isotropic shielding of 50k ‘hold-out’ atoms with a mean error of less than 1.9 ppm. For the rapid prediction of new query molecules, the models were trained on geometries from an inexpensive theory. Furthermore, by using a Δ-ML strategy, we quench the error below 1.4 ppm. Finally, we test the transferability on non-trivial benchmark sets that include benchmark molecules comprising 10–17 heavy atoms and drugs.

Improved Prediction of Carbonless NMR Spectra by the Machine Learning of Theoretical and Fragment Descriptors for Environmental Mixture Analysis

As the first multidimensional NMR approach, 2D J-resolved (2DJ) spectroscopy is distinguished by signal resolution and detection sensitivity with remarkable advantages for the exhaustive evaluation of complex mixtures and environmental samples due to its carbonless feature without the requirement of ¹³C connectivity. Generally, the 2DJ signal assignment of metabolic mixtures is problematic in spite of references to experimental NMR databases, owing to the existence of metabolic "dark matter." In this study, a new method to predict 2DJ spectra was developed with a combination of quantum mechanical (QM) computation and machine learning (ML). The predictive accuracy of J-coupling constants was evaluated using validated data. The root-mean-square deviation (RMSD) for QM computation was 3.52 Hz, while the RMSD for QM + ML was 1.21 Hz, indicating a substantial increase in predictive accuracy. The proposed model was applied to predict the 2DJ spectra of 60 standard substances and 55 components of seawater. Furthermore, two practical environmental samples were used to evaluate the robustness of the constructed predictive model. A J-coupling tree and J-split spectra produced from QM + ML of aliphatic moieties had good consistency with the experimental data, as compared with the theoretical data produced by QM computation. The predicted J-coupling tree for the J-coupling multiplet analysis of freely rotating bonds in the complex mixture, which is traditionally difficult, was interpretable. In addition, in silico identification of the J-split ¹H NMR signals, which was independent of experimental databases, aided in the discovery of new components in a mixture.

Toward Accurate Predictions of Atomic Properties via Quantum Mechanics Descriptors Augmented Graph Convolutional Neural Network: Application of This Novel Approach in NMR Chemical Shifts Predictions

In this study, an augmented Graph Convolutional Network (GCN) with quantum mechanics (QM) descriptors was reported for its accurate predictions of NMR chemical shifts with respect to experimental values. The prediction errors of ¹³C/¹H NMR chemical shifts can be as small as 2.14/0.11 ppm. There are two crucial characteristics for this modified GCN: in one aspect, such a novel neural network could efficiently extract the overall molecule structure information; in another aspect, it could accurately solve the chemical environment of the target atom. As there exists an imperfect linear regression between the experimental NMR chemical shifts (δ) and the density functional theory (DFT) calculated isotropic shielding constants (σ), the inclusion of QM descriptors within GCN can largely improve its performance. Moreover, few-shot learning also becomes feasible with these descriptors. The success of this novel GCN in chemical shifts predictions also indicates its potential applicability for other computational studies.

Accurate predictions of aqueous solubility of drug molecules via the multilevel graph convolutional network (MGCN) and SchNet architectures

Deep learning based methods have been widely applied to predict various kinds of molecular properties in the pharmaceutical industry with increasingly more success. In this study, we propose two novel models for aqueous solubility predictions, based on the Multilevel Graph Convolutional Network (MGCN) and SchNet architectures, respectively. The advantage of the MGCN lies in the fact that it could extract the graph features of the target molecules directly from the (3D) structural information; therefore, it doesn't need to rely on a lot of intra-molecular descriptors to learn the features, which are of significance for accurate predictions of the molecular properties. The SchNet performs well in modelling the interatomic interactions inside a molecule, and such a deep learning architecture is also capable of extracting structural information and further predicting the related properties. The actual accuracy of these two novel approaches was systematically benchmarked with four different independent datasets. We found that both the MGCN and SchNet models performed well for aqueous solubility predictions. In the future, we believe such promising predictive models will be applicable to enhancing the efficiency of the screening, crystallization and delivery of drug molecules, essentially as a useful tool to promote the development of molecular pharmaceutics.

Pharmaceutical Hydrates Analysis-Overview of Methods and Recent Advances

This review discusses a set of instrumental and computational methods that are used to characterize hydrated forms of APIs (active pharmaceutical ingredients). The focus has been put on highlighting advantages as well as on presenting some limitations of the selected analytical approaches. This has been performed in order to facilitate the choice of an appropriate method depending on the type of the structural feature that is to be analyzed, that is, degree of hydration, crystal structure and dynamics, and (de)hydration kinetics. The presented techniques include X-ray diffraction (single crystal X-ray diffraction (SCXRD), powder X-ray diffraction (PXRD)), spectroscopic (solid state nuclear magnetic resonance spectroscopy (ssNMR), Fourier-transformed infrared spectroscopy (FT-IR), Raman spectroscopy), thermal (differential scanning calorimetry (DSC), thermogravimetric analysis (TGA)), gravimetric (dynamic vapour sorption (DVS)), and computational (molecular mechanics (MM), Quantum Mechanics (QM), molecular dynamics (MD)) methods. Further, the successful applications of the presented methods in the studies of hydrated APIs as well as studies on the excipients' influence on these processes have been described in many examples.