From Nontargeted to Targeted Analysis: Feature Selection in the Differentiation of Truffle Species ( spp.) Using H NMR Spectroscopy and Support Vector Machine.

The price of different truffle types varies according to their culinary value, sometimes by more than a factor of 10. Nonprofessionals can hardly distinguish visually the species within the white or black truffles, making the possibility of food fraud very easy. Therefore, the identification of different truffle species ( spp.) is an analytical task that could be solved in this study. The polar extract from a total of 80 truffle samples was analyzed by H NMR spectroscopy in combination with chemometric methods covering five commercially relevant species. All classification models were validated applying a repeated nested cross-validation. In direct comparison, the two very similar looking and closely related black representatives and could be classified 100% correctly. The most expensive truffle could be distinguished 100% from the other relevant white truffle . In addition, signals for a potential and a potential marker for targeted approaches could be detected, and the corresponding molecules were identified as betaine and ribonate. A model covering all five truffle species , , , , and was able to correctly discriminate between each of the species.