Genius: a genetic algorithm for automated structure elucidation from 13C NMR spectra

The automated structure elucidation of organic molecules from experimentally obtained properties is extended by an entirely new approach. A genetic algorithm is implemented that uses molecular constitution structures as individuals. With this approach, the structure of organic molecules can be optimized to meet experimental criteria, if in addition a fast and accurate method for the prediction of the used physical or chemical features is available. This is demonstrated using 13C NMR spectrum as readily obtainable information. By means of artificial neural networks a fast and accurate method for calculating the 13C NMR spectrum of the generated structures exists. The method is implemented and tested successfully for organic molecules with up to 18 non-hydrogen atoms.     

Effective consideration of ring structures in CAST/CNMR for highly accurate C NMR chemical shift prediction

A new function that effectively takes into account ring structural environments achieves extensive highly accurate prediction of ¹³C NMR chemical shift in the CAST/CNMR system. The approach adapts a fast and flexible ring perception algorithm and a new CAST coding method for the ring information. ¹³C NMR chemical shift prediction is performed for complicated polycyclic natural products and their synthetic intermediates as the demonstration, which shows the reliability of the function in extending the scope of the practically accurate ¹³C NMR prediction for wide range of organic compounds.     

Novel methods of automated structure elucidation based on 13C NMR spectroscopy

Three new approaches for automated structure elucidations of organic molecules using NMR spectroscopic data were introduced recently. These approaches apply a neural network 13C NMR chemical shift prediction method to rank the results of structure generators by their agreement of the predicted and experimental chemical shifts. These three existing implementations are compared using realistic example molecules. The applicability and reliability of such approaches is addressed.     

Probing chemical shifts of invisible states of proteins with relaxation dispersion NMR spectroscopy: how well can we do?

Carr-Purcell-Meiboom-Gill relaxation dispersion NMR spectroscopy has evolved into a powerful approach for the study of low populated, invisible conformations of biological molecules. One of the powerful features of the experiment is that chemical shift differences between the exchanging conformers can be obtained, providing structural information about invisible excited states. Through the development of new labeling approaches and NMR experiments it is now possible to measure backbone 13C(alpha) and 13CO relaxation dispersion profiles in proteins without complications from 13C-13C couplings. Such measurements are presented here, along with those that probe exchange using 15N and 1HN nuclei. A key experimental design has been the choice of an exchanging system where excited-state chemical shifts were known from independent measurement. Thus it is possible to evaluate quantitatively the accuracy of chemical shift differences obtained in dispersion experiments and to establish that in general very accurate values can be obtained. The experimental work is supplemented by computations that suggest that similarly accurate shifts can be measured in many cases for systems with exchange rates and populations that fall within the range of those that can be quantified by relaxation dispersion. The accuracy of the extracted chemical shifts opens up the possibility of obtaining quantitative structural information of invisible states of the sort that is now available from chemical shifts recorded on ground states of proteins.     

Quantum chemical and nuclear magnetic resonance spectral studies on molecular properties and electronic structure of berberine and berberrubine

The structural and electronic properties of berberine and berberrubine have been studied extensively using density functional theory (DFT) employing B3LYP exchange correlation. The geometries of these molecules have been fully optimized at the B3LYP/6-311G** level. The chemical shift of 1H and 13C resonances in NMR spectra of these molecules have been calculated using the gauge invariant atomic model (GIAO) method as implemented in Gaussian 98. One- and two-dimensional HSQC (1H-13C), HMBC (1H-13C) and ROESY (1H-1H) spectra were recorded at 500 MHz for the berberine molecule in D(2)O solution. All proton and carbon resonances were unambiguously assigned, and inter-proton distances obtained from ten observed NOE contacts. A restrained molecular dynamics (RMD) approach was used to get the optimized solution structure of berberine. The structure of berberine and berberrubine molecules was also obtained using the ROESY data available in literature. Comparison of the calculated NMR chemical shifts with the experimental values revealed that DFT methods produce very good results for both proton and carbon chemical shifts. The importance of the basis sets to the calculated NMR parameters is discussed. It has been found that calculated structure and chemical shifts in the gas phase predicted with B3LYP/6-311G** are in very good agreement with the present experimental data and the measured values reported earlier.     

G-matrix Fourier transform NOESY-based protocol for high-quality protein structure determination

A protocol for high-quality structure determination based on G-matrix Fourier transform (GFT) NMR is presented. Five through-bond chemical shift correlation experiments providing 4D and 5D spectral information at high digital resolution are performed for resonance assignment. These are combined with a newly implemented (4,3)D GFT NOESY experiment which encodes information of 4D 15N/15N-, 13C(alipahtic)/15N-, and 13C(aliphatic)/13C(aliphatic)-resolved [1H,1H]-NOESY in two subspectra, each containing one component of chemical shift doublets arising from 4D --> 3D projection at omega1:Omega(1H) +/- Omega(X) [X = 15N,13C(aliphatic)]. The peaks located at the centers of the doublets are obtained from simultaneous 3D 15N/13C(aliphatic)/13C(aromatic)-resolved [1H,1H]-NOESY, wherein NOEs detected on aromatic protons are also obtained. The protocol was applied for determining a high-quality structure of the 14 kDa Northeast Structural Genomics consortium target protein, YqfB (PDB ID ). Through-bond correlation and NOESY spectra were acquired, respectively, in 16.9 and 39 h (30 h for shift doublets, 9 h for central peaks) on a 600 MHz spectrometer equipped with a cryogenic probe. The rapidly collected highly resolved 4D NOESY information allows one to assign the majority of NOEs directly from chemical shifts, which yields accurate initial structures "within" approximately 2 angstroms of the final structure. Information theoretical "QUEEN" analysis of initial distance limit constraint networks revealed that, in contrast to structure-based protocols, such NOE assignment is not biased toward identifying additional constraints that tend to be redundant with respect to the available constraint network. The protocol enables rapid NMR data collection for robust high-quality structure determination of proteins up to approximately 20-25 kDa in high-throughput.     

Predicting chemical shifts with graph neural networks

Inferring molecular structure from Nuclear Magnetic Resonance (NMR) measurements requires an accurate forward model that can predict chemical shifts from 3D structure. Current forward models are limited to specific molecules like proteins and state-of-the-art models are not differentiable. Thus they cannot be used with gradient methods like biased molecular dynamics. Here we use graph neural networks (GNNs) for NMR chemical shift prediction. Our GNN can model chemical shifts accurately and capture important phenomena like hydrogen bonding induced downfield shift between multiple proteins, secondary structure effects, and predict shifts of organic molecules. Previous empirical NMR models of protein NMR have relied on careful feature engineering with domain expertise. These GNNs are trained from data alone with no feature engineering yet are as accurate and can work on arbitrary molecular structures. The models are also efficient, able to compute one million chemical shifts in about 5 seconds. This work enables a new category of NMR models that have multiple interacting types of macromolecules.

This model can predict chemical shifts on proteins and small molecules purely from atom elements and coordinates. It can capture important phenomena like hydrogen bonding induced downfield shift, thus can be used to infer intermolecular interactions.     

Algorithm for advanced canonical coding of planar chemical structures that considers stereochemical and symmetric information

We describe a rigorous and fast algorithm for advanced canonical coding of planar chemical structures based on the algorithm of Faulon et al. (J. Chem. Inf. Comput. Sci. 2004, 44, 427-436). Our algorithm works well even for highly symmetric structures; moreover, an advantage of our algorithm includes providing a rigorous canonical numbering of atoms with a consideration of stereochemistry and recognizing symmetric moieties. The planar structural line notation with the canonical numbering is also fit for use with stereochemical line notation. These capabilities are usable for general purposes in chemical structural coding and are particularly essential for detecting equivalent atoms in NMR studies. This algorithm was implemented on a 13C NMR chemical shift prediction system CAST/CNMR. Applications of the algorithm to several organic compounds demonstrate the practical efficiency of the rigorous coding.     

The development of an NMR chemical shift prediction application with the accuracy necessary to grade proton NMR spectra for identity

We have developed an NMR chemical shift prediction system that enables high throughput automatic grading of NMR spectra. In support of high throughput synthetic efforts for our drug discovery program, a rapid and accurate analysis for identity was needed. The system was designed and implemented to take advantage of the NMR assignments that had been tabulated on internally generated research compounds. The system has been operational for four years and has been used in conjunction with an internally written grading program to successfully analyze several hundred thousand samples based only on their 1D ¹H spectrum. A focused test of the system's accuracy on 1006 molecules demonstrated the ability to estimate the proton chemical shift with an average error of +/?0.16 ppm. This level of chemical shift accuracy allows for reliable structure confirmation by automated analysis using only proton NMR. Copyright © 2009 John Wiley & Sons, Ltd.     

Solid-state NMR and quantum chemical investigations of 13Calpha shielding tensor magnitudes and orientations in peptides: determining phi and psi torsion angles

We report the experimental determination of the (13)C(alpha) chemical shift tensors of Ala, Leu, Val, Phe, and Met in a number of polycrystalline peptides with known X-ray or de novo solid-state NMR structures. The 700 Hz dipolar coupling between (13)C(alpha) and its directly bonded (14)N permits extraction of both the magnitude and the orientation of the shielding tensor with respect to the C(alpha)-N bond vector. The chemical shift anisotropy (CSA) is recoupled under magic-angle spinning using the SUPER technique (Liu et al., J. Magn. Reson. 2002, 155, 15-28) to yield quasi-static chemical shift powder patterns. The tensor orientation is extracted from the (13)C-(14)N dipolar modulation of the powder line shapes. The magnitudes and orientations of the experimental (13)C(alpha) chemical shift tensors are found to be in good accord with those predicted from quantum chemical calculations. Using these principal values and orientations, supplemented with previously measured tensor orientations from (13)C-(15)N and (13)C-(1)H dipolar experiments, we are able to predict the (phi, psi, chi(1)) angles of Ala and Val within 5.8 degrees of the crystallographic values. This opens up a route to accurate determination of torsion angles in proteins based on shielding tensor magnitude and orientation information using labeled compounds, as well as the structure elucidation of noncrystalline organic compounds using natural abundance (13)C NMR techniques.     

CAST/CNMR: highly accurate C NMR chemical shift prediction system considering stereochemistry

Accurate, practical prediction of ¹³C NMR chemical shifts has been achieved with a new system, CAST/CNMR, taking account of stereochemistry. The CAST/CNMR system has solved the critical problem of the accurate distinction of differences and similarities in stereochemical structures around a specific carbon, which has not yet been achieved by any other database-oriented system for prediction of ¹³C NMR chemical shifts. CAST/CNMR uses a three-dimensional structural database together with a ¹³C NMR spectral database. Absolute/relative configurational and conformational structural information are described by the CAST (CAnonical-representation of STereochemistry) coding method. This paper provides an overview of the CAST/CNMR system, and describes its application to two natural products as examples.     

Structure validation of natural products by quantum-mechanical GIAO calculations of 13C NMR chemical shifts

Geometry optimization and GIAO (gauge including atomic orbitals) (13)C NMR chemical shift calculations at Hartree-Fock level, using the 6-31G(d) basis set, are proposed as a tool to be applied in the structural characterization of new organic compounds, thus providing useful support in the interpretation of experimental NMR data. Parameters related to linear correlation plots of computed versus experimental (13)C NMR chemical shifts for fourteen low-polar natural products, containing 10-20 carbon atoms, were employed to assess the reliability of the proposed structures. A comparison with the hybrid B3LYP method was carried out to evaluate electron correlation contributions to the calculation of (13)C NMR chemical shifts and, eventually, to extend the applicability of such computational methods to the interpretation of NMR spectra in apolar solutions. The method was tested by studying three examples of revised structure assignments, analyzing how the theoretical (13)C chemical shifts of both correct and incorrect structures matched the experimental data.     

Simultaneous de novo identification of molecules in chemical mixtures by doubly indirect covariance NMR spectroscopy

The detailed characterization of complex molecular mixtures plays a key role in many areas of modern Chemistry. Here we report a novel NMR spectroscopic method that deconvolutes a complex mixture of organic molecules simultaneously into individual components and depicts their chemical structure without requiring physical separation of the components. Doubly indirect covariance spectroscopy is introduced and applied to 2D (13)C-(1)H HSQC and 2D (1)H-(1)H COSY spectra, which results in a (13)C-(13)C 2D spectrum with unprecedented high resolution. This reconstituted spectrum is indeed a carbon-connectivity map that can be directly analyzed with basic graph theory to obtain the skeletal structures of individual mixture components or their fragments. The method is demonstrated for a model mixture and a natural product mixture extracted from cancer cells. Its suitability for automation makes this approach attractive for the analysis of a broad range of mixtures of natural or synthetic products.     

Density functional theory study of 13C NMR chemical shift of chlorinated compounds

The use of the standard density functional theory (DFT) leads to an overestimation of the paramagnetic contribution and underestimation of the shielding constants, especially for chlorinated carbon nuclei. For that reason, the predictions of chlorinated compounds often yield too high chemical shift values. In this study, the WC04 functional is shown to be capable of reducing the overestimation of the chemical shift of Cl‐bonded carbons in standard DFT functionals and to show a good performance in the prediction of ¹³C NMR chemical shifts of chlorinated organic compounds. The capability is attributed to the minimization of the contributions that intensively increase the chemical shift in the WC04. Extensive computations and analyses were performed to search for the optimal procedure for WC04. The B3LYP and mPW1PW91 standard functionals were also used to evaluate the performance. Through detailed comparisons between the basis set effects and the solvent effects on the results, the gas‐phase GIAO/WC04/6‐311+G(2d,p)//B3LYP/6‐31+G(d,p) was found to be specifically suitable for the prediction of ¹³C NMR chemical shifts of chlorides in both chlorinated and non‐chlorinated carbons. Further tests with eight molecules in the probe set sufficiently confirmed that WC04 was undoubtedly effective for accurately predicting ¹³C NMR chemical shifts of chlorinated organic compounds. Copyright © 2012 John Wiley & Sons, Ltd.     

2D NMR方法研究抗癌药物冬凌草乙素的结构与谱线归属

用异核化学位移相关谱、远程异核化学位移相关谱和同核化学位移相关谱等现代核磁共振技术对抗癌中草药冬凌草中分离出的抗癌、抗菌有效成分冬凌草乙素分子的~(13)C和~1H化学位移进行了完全归属,为冬凌草乙素分子溶液中的三维空间结构研究提供了可靠的结构参数。     

Simple method for identification of skeletons of aporphine alkaloids from 13C NMR data using artificial neural networks

This paper describes the use of artificial neural networks as a theoretical tool in the structural determination of alkaloids from (13)C NMR chemical shift data, aiming to identify skeletal types of those compounds. For that, 162 aporphine alkaloids belonging to 12 different skeletons were codified with their respective (13)C NMR chemical shifts. Each skeleton pertaining to aporphine alkaloid type was used as output, and the (13)C NMR chemical shifts were used as input data of the net. Analyzing the obtained results, one can then affirm the skeleton to which each one of these compounds belongs with high degree of confidence (over 97%). The relation between the correlation coefficient and the number of epochs and the architecture of net (3-layer MLP or 4-layer MLP) were analyzed, too. The analysis showed that the results predicted by the 3-layer MLP networks trained with a number of the epochs higher than 900 epochs are the best ones. The artificial neural nets were shown to be a simple and efficient tool to solve structural elucidation problems making use of (13)C NMR chemical shift data, even when a similarity between the searched skeletons occurs, offering fast and accurate results to identification of skeletons of organic compounds.     

Three-dimensional solution NMR spectroscopy of complex structures and mixtures

This paper reviews the non-biological applications of three dimensional NMR (3D-NMR) spectroscopy methodologies for studying chemical structures in polymer science, dendrimer research, organometallic chemistry, organosilicon chemistry, and mixtures of small organic molecules. Four methodologies for solving chemical structure problems are described, where the appropriate method is determined by the presence or absence of a third X nucleus (in addition to (1)H and (13)C) with suitable NMR properties.     

Application of unsymmetrical indirect covariance NMR methods to the computation of the (13)C <--> (15)N HSQC-IMPEACH and (13)C <--> (15)N HMBC-IMPEACH correlation spectra

Utilization of long-range (1)H--(15)N heteronuclear chemical shift correlation has continually grown in importance since the first applications were reported in 1995. More recently, indirect covariance NMR methods have been introduced followed by the development of unsymmetrical indirect covariance processing methods. The latter technique has been shown to allow the calculation of hyphenated 2D NMR data matrices from more readily acquired nonhyphenated 2D NMR spectra. We recently reported the use of unsymmetrical indirect covariance processing to combine (1)H--(13)C GHSQC and (1)H--(15)N GHMBC long-range spectra to yield a (13)C--(15)N HSQC-HMBC chemical shift correlation spectrum that could not be acquired in a reasonable period of time without resorting to (15)N-labeled molecules. We now report the unsymmetrical indirect covariance processing of (1)H--(13)C GHMBC and (1)H--(15)N IMPEACH spectra to afford a (13)C--(15)N HMBC-IMPEACH spectrum that has the potential to span as many as six to eight bonds. Correlations for carbon resonances long-range coupled to a protonated carbon in the (1)H--(13)C HMBC spectrum are transferred via the long-range (1)H--(15)N coupling pathway in the (1)H--(15)N IMPEACH spectrum to afford a much broader range of correlation possibilities in the (13)C--(15)N HMBC-IMPEACH correlation spectrum. The indole alkaloid vincamine is used as a model compound to illustrate the application of the method.