Variety identification of wheat using mass spectrometry with neural networks and the influence of mass spectra processing prior to neural network analysis

Helle Aagaard Sørensen, Maria Maddalena Sperotto, M. Petersen, Can Kesmir, Louise Radzikowski, Susanne Jacobsen, Ib Søndergaard

    Research output: Contribution to journalJournal articleResearchpeer-review


    The performance of matrix-assisted laser desorption/ionisation time-of-flight mass spectrometry with neural networks in wheat variety classification is further evaluated.(1) Two principal issues were studied: (a) the number of varieties that could be classified correctly; and (b) various means of preprocessing mass spectrometric data. The number of wheat varieties tested was increased from 10 to 30. The main pre-processing method investigated was based on Gaussian smoothing of the spectra, but other methods based on normalisation procedures and multiplicative scatter correction of data were also used. With the final method, it was possible to classify 30 wheat varieties with 87% correctly classified mass spectra and a correlation coefficient of 0.90.
    Original languageEnglish
    JournalRapid Communications in Mass Spectrometry
    Issue number12
    Pages (from-to)1232-1237
    Publication statusPublished - 2002

    Cite this