Gas Chromatography/Mass Spectrometry-Based Metabolic Profiling and Differentiation of Ginseng Roots According to Cultivation Age Using Variable Selection

Abstract

Ginseng roots are an important herbal resource worldwide, and the adulteration of ginseng with age is recognized as a serious problem. It is therefore crucial to develop objective criteria or standard protocols for differentiating ginseng root samples according to their cultivation age. The reported study used GC/MS combined with multivariate statistical analysis with variable selection to obtain metabolic profiling and an optimal partial least squares-discriminant analysis (PLS-DA) model for the differentiation of ginseng according to cultivation age. Relative levels of various metabolites, such as amino acids, alcohols, fatty acids, organic acids, and sugars, were measured for various ginseng cultivation ages. Increasing cultivation age resulted in the production of higher levels of panaxynol and panaxydol, which are active polyacetylene compounds in ginseng. In addition, optimized PLS-DA models for the prediction of ginseng age were obtained by selecting variables based on a variable importance in the projection cut-off value of 1.3. Proline, glucaric acid, mannose, gluconic acid, glucuronic acid, myoinositol, panaxydol, and panaxynol are suggested as key and relevant compounds with which to differentiate the age of ginseng samples. The findings of this study suggest that GC/MS-based metabolic profiling can be used to differentiate ginseng samples according to cultivation age.

Metabolomic Approach for Discrimination of Four- and Six-Year-Old Red Ginseng (<i>Panax ginseng</i>) Using UPLC-QToF-MS

“…We proposed one potential marker substance was isolated from six-year-old red ginseng and identified by NMR spectroscopic analyses. As previously mentioned, age discrimination of ginseng samples has been studied previously, [6][7][8][9][10][11][12] but only considered fresh ginseng; this is the first report of marker substances for the discrimination of four-and six-year-old red ginseng. These results are also helpful for quality control and standardization and to provide a scientific basis for pharmacological research of red ginseng.…”

“…The accurate determination of the cultivation age of ginseng is a very important problem in the market, although the cultivation age of ginseng can hardly be determined by the physical appearance alone, such as by the number of stem vestiges in rhizome. 5) In recent years, attempts have been made to overcome this difficulty in discriminating between the two categories of products using metabolomics by various instrumental methods such as NMR, [6][7][8] GC-MS 9,10) and LC-MS. 11,12) However, the studies regarding red ginseng were unable to produce reliable results. In this study, we focused on reliable methods to discriminate the cultivation ages of red ginseng for quality control and prevention of adulteration of the market.…”

Traceability of Boletaceae mushrooms using data fusion of UV–visible and FTIR combined with chemometrics methods

“…The performance of the models was evaluated by the RMSEE, the RMSECV and the accuracy of the training set and test set. In addition, RMSEE expresses the differences between actual and predicted, and RMSECV is used to prevent overfitting . As shown in Table , the values of RMSEE and RMSECV of UV–visible are closely similar to the values of FTIR, but the accuracy of the training set and the test set of FTIR is obviously higher than UV–visible.…”

“…RMSEE is a used to evaluate the calibration error, and RMSECV is computed based on a cross-validation that can separate calibration samples into training set and the test set. 35 Generally, the better the classification model is, the lower RMSEE and RMSECV values are.…”

“…Root mean square error of estimation (RMSEE) and root mean square error of cross‐validation (RMSECV) are two important parameters to evaluate the performance of a PLS‐DA model. RMSEE is a used to evaluate the calibration error, and RMSECV is computed based on a cross‐validation that can separate calibration samples into training set and the test set . Generally, the better the classification model is, the lower RMSEE and RMSECV values are.…”