An application of neural networks in chemistry. Prediction of13C NMR chemical shifts

Basic definitions of neural networks are given in terms of oriented graphs. Partial derivatives of an objective function with respect to the weight and threshold coefficients are derived. These derivatives are very important for the adaptation process, carried out by a version of the gradient method of We neural network considered. The stability of the adapted neural network toward small changes — perturbation — of input activities is described by sensitivities. The theory is illustrated by application of simple neural networks that reflect the topology of molecules to the classification of¹³C NMR chemical shifts of secondary carbons in acyclic alkanes.     

Structure-based predictions of H NMR chemical shifts of sesquiterpene lactones using neural networks

In this work the prediction of ¹H NMR chemical shifts of CH_n protons of sesquiterpene lactones by means of neural networks is described. This method is based on the incorporation of experimental chemical shifts of protons of sesquiterpene lactones as additional memory of an associative neural network system previously trained with chemical shifts of other organic compounds. One advantage of this method is its ability to distinguish between CH₂ diastereotopic protons belonging to rigid substructures since stereochemistry is considered. This is achieved via the automatic conversion of the 2D structure diagram into a 3D molecular structure. The predicted ¹H NMR chemical shifts of the sesquiterpene lactones showed a high level of accuracy. This is the first report on a fully automatic proton assignment of structures of sesquiterpene lactones of an accuracy that allows its use in structure elucidation.     

Computational studies of 13C NMR chemical shifts of saccharides

The (13)C NMR chemical shifts for alpha-D-lyxofuranose, alpha-D-lyxopyranose (1)C(4), alpha-D-lyxopyranose (4)C(1), alpha-D-glucopyranose (4)C(1), and alpha-D-glucofuranose have been studied at ab initio and density-functional theory levels using TZVP quality basis set. The methods were tested by calculating the nuclear magnetic shieldings for tetramethylsilane (TMS) at different levels of theory using large basis sets. Test calculations on the monosaccharides showed B3LYP(TZVP) and BP86(TZVP) to be cost-efficient levels of theory for calculation of NMR chemical shifts of carbohydrates. The accuracy of the molecular structures and chemical shifts calculated at the B3LYP(TZVP) level is comparable to those obtained at the MP2(TZVP) level. Solvent effects were considered by surrounding the saccharides by water molecules and also by employing a continuum solvent model. None of the applied methods to consider solvent effects was successful. The B3LYP(TZVP) and MP2(TZVP)(13)C NMR chemical shift calculations yielded without solvent and rovibrational corrections an average deviation of 5.4 ppm and 5.0 ppm between calculated and measured shifts. A closer agreement between calculated and measured chemical shifts can be obtained by using a reference compound that is structurally reminiscent of saccharides such as neat methanol. An accurate shielding reference for carbohydrates can be constructed by adding an empirical constant shift to the calculated chemical shifts, deduced from comparisons of B3LYP(TZVP) or BP86(TZVP) and measured chemical shifts of monosaccharides. The systematic deviation of about 3 ppm for O(1)H chemical shifts can be designed to hydrogen bonding, whereas solvent effects on the (1)H NMR chemical shifts of C(1)H were found to be small. At the B3LYP(TZVP) level, the barrier for the torsional motion of the hydroxyl group at C(6) in alpha-D-glucofuranose was calculated to 7.5 kcal mol(-1). The torsional displacement was found to introduce large changes of up to 10 ppm to the (13)C NMR chemical shifts yielding uncertainties of about +/-2 ppm in the chemical shifts.     

13C NMR chemical shifts of N-unsubstituted- and N-methyl-pyrazole derivatives

¹³C shielding data for 100 derivatives of pyrazole are reported. These include methyl, ethyl, n-propyl, tert-butyl, phenyl, hydroxymethyl, carboxyl, ethoxycarbonyl, cyano, amino, hydrazino, nitro, azido, chloro, bromo and iodo groups as substituents on the ring carbon atoms.     

Prediction of 13C NMR chemical shifts in substituted naphthalenes

The prediction of the ¹³C NMR signals for derivatives of naphthalene has been investigated using mathematical modeling techniques. Two empirical multiple regression models which utilize the field, resonance, and Charton's steric parameters together with molar refractivity were developed, one for α- and the other for β-substituted naphthalene derivatives. In the α case the model had a correlation coefficient of observed versus predicted line positions of r = 0.973 with a standard deviation of 2.2 ppm while in the β case r = 0.979 with the standard deviation being 2.3 ppm. The database consisted of 3152 signals from 394 naphthalene derivatives. We also report the use of the Taft steric parameter in place of the Charton steric parameter in the above- mentioned prediction equations. Received: 19 June 1998 / Accepted: 20 October 1998 / Published online: 16 March 1999     

13C NMR chemical shifts of some highly-branched acyclic compounds

A study has been made of the ¹³C chemical shifts of a number of acyclic alkanes, alkenes, nitriles and ketones which contain quaternary carbon atoms. Similar data have also been obtained for the series of compounds involved in the synthesis of triisopropylacetic acid. Substituent effects and steric factors in these highly substituted compounds are discussed in relation to the chemical shifts.     

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.     

13C and 15N NMR chemical shifts of alkylsubstituted benzonitriles

¹³C NMR data of 14 and ¹⁵N NMR data of four alkylsubstituted benzonitriles are reported and discussed in relation to substituent effects and stereochemistry.     

On the determination of micellar aggregation numbers from the concentration dependence of13C NMR chemical shifts

The assumptions underlying the extraction of micellar aggregation numbers by means of applying the mass action law to the concentration dependence of¹³C Nuclear Magnetic Resonance (NMR) shift data are discussed. Such data are presented for sodium dodecylsulfate and it is shown that the extracted aggregation numbers are far too small. It is argued that this is in part due to a failure of the mass action law to describe the micellization process but also due to covariance in the parameters of the mass action law. We also suggest a way to analyse¹³C shifts from surfactant systems that is void of artefacts due to changes in volume magnetic susceptibilities and other unwanted artefacts. Finally, we point out that by combining¹³C shifts with the fraction of micellized surfactant (as measured by for instance self diffusion coefficients) it should be possible to monitor changes in micellar shapes as the conditions are changed.     

丁烷衍生物类木脂素13C NMR化学位移预测

借助原子电性作用矢量(AEIV)和原子杂化状态指数(AHSI),对39种丁烷衍生物类木脂素共计854个等价C原子进行表征,并建立用于模拟该类分子13C NMR化学位移的多元线性回归方程.所得定量结构波谱关系(QSSR)模型及留一法交互检验相关系数分别为r=0.981和q=0.962.进一步用从马尾松松针中分离所得新木脂素中20个13C NMR化学位移对模型进行外部验证,预测结果与实验值较接近.表明所建模型有良好稳定性和泛化力,可对丁烷衍生物类木脂素13C NMR谱学数据准确模拟.     

Substituent effects in the 13C NMR chemical shifts of alpha-mono-substituted acetonitriles

13C chemical shifts empirical calculations, through a very simple additivity relationship, for the alpha-methylene carbon of some alpha-mono-substituted acetonitriles, Y-CH(2)-CN (Y=H, F, Cl, Br, I, OMe, OEt, SMe, SEt, NMe(2), NEt(2), Me and Et), lead to similar, or even better, results in comparison to the reported values obtained through Quantum Mechanics methods. The observed deviations, for some substituents, are very similar for both approaches. This divergence between experimental and calculated, either empirically or theoretically, values are smaller than for the corresponding acetones, amides, acetic acids and methyl esters, which had been named non-additivity effects (or intramolecular interaction chemical shifts, ICS) and attributed to some orbital interactions. Here, these orbital interactions do not seem to be the main reason for the non-additivity effects in the empirical calculations, which must be due solely to the magnetic anisotropy of the heavy atom present in the substituent. These deviations, which were also observed in the theoretical calculations, were attributed in that case to the non-inclusion of relativistic effects and spin-orbit coupling in the Hamiltonian. Some divergence is also observed for the cyano carbon chemical shifts, probably due to the same reasons.     

Representation of13C NMR chemical shifts in polychlorobromomethanes using a distance matrix for the heterosystems

On the basis of a distance matrix for heterosystems, we propose a form for the topological dependence for the carbon NMR chemical shifts in polyhalomethanes.Translated from Izvestiya Akademii Nauk SSSR, Seriya Khimicheskaya, No. 1, pp. 58–60, January, 1990.     

Application of quantum chemical calculations of C NMR chemical shifts to quinoxaline structure determination

Comparison of experimental and theoretical (GIAO DFT) ¹³C NMR chemical shifts allows the reliable assignment of isomeric structures of heteroaromatic compounds. This methodology was applied to establish the structures of isomeric quinoxalines. A modern 1D NOE technique permitted independent proof of the proposed structures.     

The13C NMR spectra of some polychlorinated dibenzo-p-dioxins. A calculation of13C chemical shifts

The¹³C NMR spectra of a number of polychiorinated dibenzo-p-dioxins (PCDD) were measured. These and previously known spectra were used for the development of a method for calculation of¹³C NMR spectra of chloroaromatics in the framework of a two-particle increment scheme for carbon chemical shifts. The scheme one allows to calculate¹³C chemical shifts for all 75 PCDD.Translated fromIzvestiya Akademii Nauk. Seriya Khimicheskaya, No. 4, pp. 760–761, April, 1994.     

Topological increment scheme for13C NMR chemical shifts in aliphatic alcohols

We propose an additive empirical scheme for calculating the¹³C NMR chemical shifts in acyclic aliphatic alcohols. We show that the electron-acceptor effect of the OH group does not damp out at the -carbon and is apparent in the shielding of the - and -carbon atoms. The effect of the OH group has two components: a constant component determined by the electron-acceptor properties of the substituent (the true a effect of the OH group, equal to the change in the chemical shift of the methane carbon upon substitution of a hydrogen atom by a hydroxyl), and a variable component of steric origin, depending on some topological characteristics of the given molecule in the immediate vicinity of the carbon atom under consideration. Analogous relations are obtained for the and effects of the OH group.Translated from Izvestiya Akademii Nauk SSSR, Seriya Khimicheskaya, No. 12, pp. 2742–2746, December, 1990.