Martin Will

Automated Structure Elucidation of Organic Molecules from 13C NMR Spectra Using Genetic Algorithms and Neural Networks

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 (13)C NMR spectrum as readily obtainable information. (13)C NMR chemical shift, intensity, and multiplicity information is available from (13)C NMR DEPT spectra. By means of artificial neural networks a fast and accurate method for calculating the (13)C NMR spectrum of the generated structures exists. The approach is limited by the size of the constitutional space that has to be searched and by the accuracy of the shift prediction for the unknown substance. The method is implemented and tested successfully for organic molecules with up to 20 non-hydrogen atoms.

Validation of structural proposals by substructure analysis and 13C NMR chemical shift prediction.

The 2D NMR-guided computer program COCON can be extremely valuable for the constitutional analysis of unknown compounds, if its results are evaluated by neural network-assisted 13C NMR chemical shift and substructure analyses. As instructive examples, data sets of four differently complex marine natural products were thoroughly investigated. As a significant step towards a true automated structure elucidation, it is shown that the primary COCON output can be safely diminished to less than 1% of its original size without losing the correct structural proposal.

Neural Network Prediction of 13C NMR Chemical Shifts of Substituted Benzenes

Summary. A multi-layer feedforward neural network was used for the prediction and assignment of 13C NMR chemical shifts of substituted benzenes. The back-propagation neural network was trained by supervised learning with the chemical shift values of about 1000 substituted benzenes from literature. The average uncertainty for the prediction of the 13C chemical shifts is as low as 1.1 ppm. In comparison to common incremental methods, essentially better results were obtained for highly substituted systems with interacting substituents.Zusammenfassung. Es wird eine Methode für die Berechnung der 13C-NMR-chemischen Verschiebungen von aromatischen Kohlenstoffatomen in substituierten Benzolen vorgestellt. Hierfür kam ein mehrschichtiges neuronales Netz mit Fehlerrückführung zum Einsatz, welches mit den Literaturwerten der chemischen Verschiebungen von über 1000 monosubstituierten Aromaten trainiert wurde. Das neuronale Netz ist in der Lage, die 13C-chemischen Verschiebungen in Aromaten unabhängig von der Anzahl ihrer Substituenten genau vorherzusagen. Die durchschnittlichen Abweichungen zu den experimentellen Werten sind kleiner als 1.1 ppm. Die Methode ist insbesondere für die Berechnung der Verschiebungswerte höhersubstituierter Benzole deutlich besser geeignet als die bekannten Inkrementverfahren, was an mehreren Beispielen gezeigt wird.