Quality Evaluation of Phellodendri Chinensis Cortex by Fingerprint–Chemical Pattern Recognition

Abstract

Phellodendri Chinensis Cortex (PCC) and Phellodendri Amurensis Cortex (PAC) are increasingly being used as traditional herbal medicines, but they are often mistaken for each other. In this study, the fingerprints of PCC from six different geographical sources were obtained by high-performance liquid chromatography, and multivariate chemometric methods were used for comprehensive analysis. Two unsupervised pattern recognition models (principal component analysis and hierarchical cluster analysis) and a supervised pattern recognition model (partial least squares discriminant analysis) were established on the basis of the chemical composition and physical traits of PCC and PAC. PCC and PAC were found to be distinguishable by these methods. The PCC category was divisible into two categories, one with more crude cork and a maximum thickness of ~1.5 mm, and the other with less net crude cork and a maximum thickness of 0.5 mm. According to the model established by partial least squares discriminant analysis (PLS-DA), the important chemical marker berberine hydrochloride was obtained and analyzed quantitatively. From these results combined with chemometric and content analyses, the preliminary classification standards for phellodendron were established as three grades: superior, first-order and mixed. Compared with the traditional identification methods of thin layer chromatography identification and microscopic identification, our method for quality evaluation is relatively simple. It provides a basis and reference for identification of PCC and enables establishment of grade standards. It also could be applied in quality control for compound preparations containing PCC.

Quality evaluation of Panax notoginseng using high‐performance liquid chromatography with chemical pattern recognition

“…Principal components (PCs) were selected from the obtained eigenvalues, the first three PCs accounted for 76.9% of the total variance, which were enough to describe the variability. The cross‐validation coefficient was Q 2 (cum) = 0.415, which indicated that the model was reasonable and had good analytical and predictive ability [28]. The 3D score plot is shown in Figure 3A.…”

“…As a result, at a confidence level of 95%, the cumulative contribution rates were R 2 X (cum) = 0.731 and R 2 Y (cum) = 0.882, the cross‐validation Q 2 (cum) in the OPLS‐DA model was 0.719. Thus, compared with PCA, it could explain most of the variance of the data and has a greater fitting and predictive ability [28]. The 2D score plot is shown in Figure 3B, from which the separation of CQ and DQ can be clearly observed.…”

High‐performance liquid chromatography–based fingerprint analysis with chemical pattern recognition for evaluation of Mahonia bealei (Fort.) Carr

“…In HCA, a dendrogram was obtained to characterize the classification results of the different sample sets by Ward's linkage as a cluster method. PCA was used to analyze and describe the distribution of the studied sample as it could explain most of data variance with a greater fitting and predictive ability [19]. Besides, PLS-DA can maximize the difference among the groups and screen out key markers that could be responsible for sample variation [21,22].…”

“…PLS-DA is a supervised CPR method to optimize classification and to screen out key markers responsible for sample variation. Many researchers have reported the quality assessment of TCMs with distinct classification of test samples on the basis of their different components using PLS-DA after comparing SA, HCA, and PCA [19,24,25]. In the PLS-DA model, R 2 X (cum) and R 2 Y (cum) are two parameters that are often calculated by the cross-validation procedure to evaluate the goodness of fit, while Q 2 (cum) is a parameter used to describe the validity of the model.…”

“…Thus, in order to simplify complex data and find hidden information behind the fingerprints, chemical pattern recognition (CPR) has been introduced to effectively and comprehensively evaluate the quality of TCMs. Different CPR models have already been used for this purpose such as hierarchical cluster analysis (HCA), principal component analysis (PCA), and partial least squares discriminant analysis (PLS-DA) [19].…”