ABSTRACT
The use of proton NMR spectroscopy allows the analysis of complex multi-component mixtures such as plant extracts by simultaneous quantification of all proton-bearing compounds and consequently all relevant substance classes. Since the spectra obtained are too complicated to be analysed visually, the classification of spectra was carried out using multivariate statistical methods. The spectroscopic data of various extracts of St. John's wort (Hypericum perforatum) samples derived from 4 different accessions extracted with 6 distinct solvents were chemometrically evaluated and calibrated using the partial least square (PLS) algorithm. In a first approach, we found a consistent correlation for the spectroscopic pattern of the extracts and the corresponding IC (50) values derived from non-selective binding to opioid receptors. Consequently, the multivariate data analysis was used to predict the pharmacological efficacy of further St. John's wort extracts on the basis of their proton NMR spectra. In a second approach a PLS 2 model was used to predict the biological activity for eight St. John's wort extracts based on two pharmacological data sets: (i) non-selective binding to opioid receptors and (ii) antagonist effect at corticotrophin-releasing factor type 1 (CRF (1)) receptors. The PLS 2 model confirmed the useful application of the presented approach to assess the quality of medicinal herbs and extracts by spectroscopic analysis derived from bioactivity-related quality parameters.