Characterization of strawberry varieties by SPME–GC–MS and Kohonen self-organizing map

Abstract

Strawberry varieties like other fruit species may be characterized from their chemical composition and more precisely, from the chemical composition of their volatiles. In this field, solid-phase micro-extraction (SPME) in the headspace mode (HS), coupled with gas chromatography–mass spectrometry (GC–MS), is known to provide a «chemical signature» representative of this volatile composition. In addition, the so-called Kohonen self-organizing maps (SOM) revealed a relevant and efficient application of artificial neural network (ANN) for reading and recognizing these signatures independently of the chemical variability characterizing biological samples. In a former experiment, 17 varieties had been fully discriminated from treatment of 70 chemical signatures proving that «chemical distances» as calculated by the SOM algorithm were much larger between varieties than between samples of the same variety. As an extension and by gambling on the size of the corresponding trained maps, it was possible to increase the resolution effect of the data treatment and to compare the variability levels of these signatures when varying some parameters like the growing year, the place of production and the fresh and frozen state of samples when analyzed. For each of these parameters, pattern features appeared and allowed a discussion according to the relative levels of the observed discrimination.