A fast multi-source information fusion strategy based on FTIR spectroscopy for geographical authentication of wild Gentiana rigescens

Gentiana rigescens Franch. ex Hemsl has long been used as a phytotherapeutic agent in various ethno-medical systems of East Asia, especially in China. However, the quality of medicinal materials varies from different geographical regions, so it is necessary to establish a fast and reliable geographical authentication method. In our study, roots, stems and leaves of G. rigescens from eight different origins (including 873 individuals) of southwest China were analyzed by Fourier transform infrared (FTIR) spectra combined with pattern recognition technology, including principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) methods. Two information fusion methods combined with two data mining methods were used to further im prove the classification performance. The results showed that PCA could hardly distinguish these origins. PLS-DA model was more suitable for the geographical origins classification. The classification performance of PLS-DA model based on the leaf FTIR matrix was better than those based on root datasets, which indicated that leaf tissues could be employed in the identification of geographical origins of G. rigescens effectively. In addition, the full spectral information derived from the fusion of infrared spectra in different regions showed the best ability to identify the geographical origins of wild G. rigescens, which had the potential to be a promising analytical tool for rapidly geographical authentication of more valuable herbal medicines.