|Sun,, Xiaobo - Ohio University|
|Cook, Shannon - Ohio University|
|Jackson, Glen - Ohio University|
|Harnly, James - Jim|
|Harrington, Peter - Ohio University|
Submitted to: Analytical Chemistry
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/13/2012
Publication Date: 3/13/2013
Citation: Sun,, X., Chen, P., Cook, S., Jackson, G., Harnly, J.M., Harrington, P. 2013. Classification of cultivation locations of Panax quinquefolius L samples using high performance liquid chromatography-electrospray ionization mass spectrometry and chemometric analysis. Analytical Chemistry. 84:3628-3634.
Interpretive Summary: Panax quinquefolius L (P. quinquefolius L) samples grown in the United States and China were analyzed using high performance liquid chromatography-mass spectrometry (HPLC—MS). The data collected from the two groups of samples were analyzed using sophisticated statistical tools, such as principal component analysis (PCA), projected difference resolution (PDR) metrics, and a fuzzy rule-building expert system (FuRES). By using these tools, P. quinquefolius L samples can be discriminated. This method has the potential to be used as an authentication method for P. quinquefolius L grown in China and the United States.
Technical Abstract: Panax quinquefolius L (P. quinquefolius L) samples grown in the United States and China were analyzed with high performance liquid chromatography-mass spectrometry (HPLC—MS). Prior to classification, the two-way datasets were subjected to pretreatment including baseline correction and retention time (RT) alignment. Principal component analysis (PCA) and projected difference resolution (PDR) metrics were used to evaluate the data quality and the pretreatment effects. A fuzzy rule-building expert system (FuRES) classifier was used to classify the P. quinquefolius L samples grown in the United States and China with the optimized partial least squares (o-PLS) classifier as the positively biased control method. A classification rate as high as 99.6 ± 0.9% with FuRES was obtained after baseline correction and RT alignment, which is equivalent to the result obtained by using the positively biased o-PLS control method (99 ± 1%). The comparison of these two types of classifiers gave a t-test of 1 with a p-value of 34%, demonstrating that there is no significant difference between the positively biased reference method and the fuzzy rule-building expert system. Both the data pretreatment methods, baseline correction and RT alignment, improved the classification rates for both FuRES and o-PLS classifiers, while RT alignment gave 19.6 and 16 percent improvement for the FuRES classification rate with and without baseline correction, respectively. From the rule obtained to classify the P. quinquefolius L samples grown in the United States and China, the first ten largest peaks with weight coefficients in decreasing importance were at RT of 14.9 min and m/z of 1031, RT of 12.6 min and m/z of 574.3, RT of 7.9 min and m/z of 595.3, RT of 10.8 min and m/z of 955.5, RT of 20.2 min and m/z of 553.3, RT of 20.2 min and m/z of 1077.8, RT of 22.6 min and m/z of 574.3, RT of 13.8 min and m/z of 1031, RT of 13 min and m/z of 1164, RT of 10.9 min and m/z of 926.1. All the peaks mentioned above had a ratio of average peak intensity to its standard deviation among 30 bootstrapped Latin partition models greater than 3. Comparisons were made among results obtained by using the whole dataset, only the first ten largest peaks in the rule and the first two largest magnitude peaks in the two-way rule. Among all the peaks in the two-way rule, the peaks with m/z of 1031 had the largest coefficient and the corresponding mass spectral peaks in two types of ginseng samples had distinctively different relative intensities. Another important feature is the intensity ratio of peak m/z 574.3 to peak m/z 1031. The average intensity ratios of peaks with m/z of 574.3 and m/z of 1031 were 1.1:1 for the P. quinquefolius L samples grown in the United State and 2.8:1 for samples grown in China, respectively. When ionization conditions change, peak intensities of m/z 1149.6 and m/z 1193.6 can be taken into consideration because m/z 1193.6 is the precursor of m/z 1149.6 and m/z 574.3 is the doubly charged species of m/z 1149.6. These peaks can be prospective biomarkers for differentiating samples from different growth regions. By using these biomarkers, P. quinquefolius samples can be discriminated preliminarily according to the two-way profiles and then chemometric analysis can be used to confirm their identities. This method has the potential to be used as an authentication method for P. quinquefolius L grown in China and the United States.