Search

The efficacy of neural network (NN) and partial least-squares (PLS) methods is compared for the prediction of NMR chemical shifts for both 1H and 13C nuclei using very large databases containing millions of chemical shifts. The chemical structure description scheme used in this work is based on individual atoms rather than functional groups. The performances of each of the methods were optimized in a systematic manner described in this work. Both of the methods, least-squares and neural network analyses, produce results of a very similar quality, but the least-squares algorithm is approximately 2--3 times faster.

Contemporary Computer-Aided Structure Elucidation (CASE) systems are heavily based on the utilization of 2D NMR spectra. The utilization of HMBC/GHMBC and COSY/GCOSY correlations generally assumes that these correlations result from (2-3)JCH and (2-3)JHH spin-spin couplings, respectively, and consequently these values are used as the default setting in these systems. Our previous studies1,2 have shown that about half of the problems studied actually contain some correlations of 4-6 bonds, so-called "nonstandard" correlations. In such cases the initial 2D NMR data are contradictory, and the correct solution is therefore not directly attainable. Unfortunately nonstandard correlations and the number of intervening bonds usually cannot be identified experimentally. In this work we suggest a new approach that we term Fuzzy Structure Generation. This allows the solution of structural problems whose 2D NMR data contain an unknown number of nonstandard correlations having different and unknown lengths. Suggested methods for the application of Fuzzy Structure Generation are described, and their application is illustrated by a series of real-world examples. We conclude that Fuzzy Structure Generation is efficient, and there is no real alternative at present in terms of a universal practical method for the structure elucidation of organic molecules from 2D NMR data.

Expert systems for spectroscopic molecular structure elucidation have been developed since the mid-1960s. Algorithms associated with the structure generation process within these systems are deterministic; that is, they are based on graph theory and combinatorial analysis. A series of expert systems utilizing 2D NMR spectra have been described in the literature and are capable of determining the molecular structures of large organic molecules including complex natural products. Recently, an opinion was expressed in the literature that these systems would fail when elucidating structures containing more than 30 heavy atoms. A suggestion was put forward that stochastic algorithms for structure generation would be necessary to overcome this shortcoming. In this article, we describe a comprehensive investigation of the capabilities of the deterministic expert system Structure Elucidator. The results of performing the structure elucidation of 250 complex natural products with this program were studied and generalized. The conclusion is that 2D NMR deterministic expert systems are certainly capable of elucidating large structures (up to about 100 heavy atoms) and can deal with the complexities associated with both poor and contradictory spectral data.

The elucidation of chemical structures from 2D NMR data commonly utilizes a combination of COSY, HMQC/HSQC, and HMBC data. Generally COSY connectivities are assumed to mostly describe the separation of protons that are separated by 1 skeletal bond (3JHH), while HMBC connectivities represent protons separated from carbon atoms by 1 to 2 skeletal bonds (2JCH and 3JCH). Obviously COSY and HMBC connectivities of lengths greater than those described have been detected. Though experimental techniques have recently been described to aid in the identification of the nature of the couplings the detection of whether a coupling is 2-bond or greater still remains a challenge in most laboratories. In the StrucEluc software system the common lengths of the connectivities, 1-bond for COSY and 1- or 2-bond for HMBC, derived from 2D NMR data are set as the default. Therefore, in the presence of any extended connectivities contradictions can appear in the 2D NMR data. In this article, algorithmic methods for the detection and removal of contradictions in 2D NMR data that have been developed in support of StrucEluc are described. The methods are based on the analysis of molecular connectivity diagrams, MCDs. These methods have been implemented in the StrucEluc system and tested by solving 50 structural problems with 2D NMR spectral data containing contradictions. The presence of contradictions was detected by the algorithm in 90% of the cases, and the contradictions were automatically removed in approximately 50% of the problems. A method of "fuzzy" structure generation in the presence of contradictions has been suggested and successfully tested in this work. This work will demonstrate examples of the application of developed methods to a number of structural problems., (Copyright 2004 American Chemical Society)

StrucEluc is an expert system that allows the computer-assisted elucidation of chemical structures based on the inputs of a series of spectral data including 1D and 2D NMR and mass spectra. The system has been enabled to allow a chemist to utilize fragments stored in a fragment database as well as user-defined fragments submitted by the chemist in the structure elucidation process. The association of fragments in this way has been shown to dramatically speed up the process of structure generation from 2D NMR data and has helped to minimize or eliminate the need for user intervention thereby further enabling the vision of automated elucidation. The use of fragments has frequently transformed very difficult 2D NMR elucidation challenges into easily solvable tasks. A strategy to utilize molecular fragments has been developed and optimized based on specific challenging examples. This strategy will be described here using real world examples. Experience gained by solving more than 150 structure elucidation problems from a variety of literature sources is also reviewed in this work.