Classification of mass spectra: A comparison of yes/no classification methods for the recognition of simple structural properties

https://doi.org/10.1016/0169-7439(94)85018-6Get rights and content

Abstract

Mass spectral classifiers for 54 structural properties have been developed using a library of 90 000 spectra and applying four complementary classification methods (k-nearest neighbor, linear discriminant analysis, SIMCA, and a neural network). Neural networks yielded best results; however, only a few structural properties could be classified with a sufficiently high predictive ability. The main difficulties to be solved for future work are the construction of a library with high-quality spectra, and the definition of structural properties that are actually reflected in low resolution mass spectra.

Werther, W., Lohninger, H., Stancl, F. and Varmuza, K., 1994. Classification of mass spectra. A comparison of yes/no classification methods for the recognition of simple structural properties. Chemometrics and Intelligent Laboratory Systems, 22: 63–76.

References (37)

  • P.C. Jurs et al.

    Computerized learning machines applied to chemical problems. Molecular formula determination from low resolution mass spectrometry

    Analytical Chemistry

    (1969)
  • J.R. Chapman
    (1978)
  • K. Varmuza
    (1980)
  • H. Lohninger et al.

    Selective detection of classes of chemical compounds by gas chromatography/mass spectrometry/pattern recognition: Polycyclic aromatic hydrocarbons and alkanes

    Analytical Chemistry

    (1987)
  • W. Werther et al.

    Development of chemometric detectors for gas chromatography/mass spectrometry by the use of a linear mapping method

  • W. Werther et al.

    EDAS-MS: Exploratory data analysis of mass spectra

  • K. Varmuza et al.

    Computer-aided interpretation of mass spectra by a combination of library search with principal component analysis

    Rapid Communications in Mass Spectrometry

    (1990)
  • V. Varmuza et al.

    A software tool for recognizing hidden relationships between mass spectra and chemical structures

    Spectroscopy International

    (1991)
  • Cited by (46)

    • Mass spectral databases for LC/MS- and GC/MS-based metabolomics: State of the field and future prospects

      2016, TrAC - Trends in Analytical Chemistry
      Citation Excerpt :

      AMDIS has been designed by NIST to deconvolute and reconstruct “pure component” spectra from complex GC/MS total ion current chromatograms. Deconvoluted mass spectra are used for compound identification via NIST database matching, while substructure interpretation is also available (complementary to the MS Search) using the classification of Varmuza et al. [73,74]. Although AMDIS is freely available, the code is not.

    • Support vector machines and its applications in chemistry

      2009, Chemometrics and Intelligent Laboratory Systems
    • Feature selection by genetic algorithms for mass spectral classifiers

      2001, Analytica Chimica Acta
      Citation Excerpt :

      In early work with mass spectral classifiers, mostly the peak intensities have been used as features. It has been shown [18,24] that an appropriate transformation of the peak data into a set of new spectral features improves the relationships between features and chemical structure information. In this work, a set of 400 spectral features has been used as summarised in Table 2; details have been described elsewhere [18,20].

    View all citing articles on Scopus
    View full text