Chapter 17. Interpretation of 13C NMR Spectra by Artificial Neural Network Techniques (NeuroCarb)

  1. Claus-Wilhelm von der Lieth2,
  2. Thomas Lütteke3 and
  3. Martin Frank2
  1. Andreas Stoeckli,
  2. Matthias Studer,
  3. Brian Cutting and
  4. Beat Ernst

Published Online: 13 NOV 2009

DOI: 10.1002/9780470029619.ch17

Bioinformatics for Glycobiology and Glycomics: An Introduction

Bioinformatics for Glycobiology and Glycomics: An Introduction

How to Cite

Stoeckli, A., Studer, M., Cutting, B. and Ernst, B. (2009) Interpretation of 13C NMR Spectra by Artificial Neural Network Techniques (NeuroCarb), in Bioinformatics for Glycobiology and Glycomics: An Introduction (eds C.-W. von der Lieth, T. Lütteke and M. Frank), John Wiley & Sons, Ltd, Chichester, UK. doi: 10.1002/9780470029619.ch17

Editor Information

  1. 2

    Molecular Structure Analysis Core Facility, Deutsches Krebsforschungszentrum (German Cancer Research Center), Heidelberg, Germany

  2. 3

    Faculty of Veterinary Medicine, Institute of Biochemistry and Endocrinology, Justus-Liebig University Gießen, Gießen, Germany

Author Information

  1. Institute of Molecular Pharmacy, Pharmacenter of the University of Basel, 4056 Basel, Switzerland

Publication History

  1. Published Online: 13 NOV 2009
  2. Published Print: 11 DEC 2009

ISBN Information

Print ISBN: 9780470016671

Online ISBN: 9780470029619

SEARCH

Keywords:

  • Neural network;
  • NeuroCarb;
  • oligosaccharide structure;
  • 13C NMR;
  • glycosylation analysis

Summary

An increasing number of biopharmaceuticals are glycoproteins, for which the correct glycosylation pattern is of critical importance for both their structure and function. One approach in the analysis of glycosylation patterns is the use of 13C NMR spectroscopy. Since the chemical shifts of adjacent monosaccharide moieties are only moderately influenced by one another, the analysis of the glycosylation pattern through 13C NMR spectroscopy is promising.

We describe an approach using artificial neural networks to simplify and accelerate the assignment of 13C NMR spectra. Our approach, entitled NeuroCarb, applies both unsupervised and supervised learning. Unsupervised learning, specifically Kohonen networks were used to categorize the test compounds according to their anomeric configuration, their group affiliation and linkage positions. Supervised learning, specifically back-propagation networks were used in later phases, both to test and apply the approach. NeuroCarb was presented the 13C NMR spectra of 65 unknown disaccharides and successfully identified 83% of the individual monosaccharides.