Harnly James, Chen Pei, Harrington Peter De B
J AOAC Int. 2013 Nov-Dec;96(6):1258-65. doi: 10.5740/jaoacint.13-290.
The AOAC INTERNATIONAL guidelines for validation of botanical identification methods were applied to the detection of Asian Ginseng [Panax ginseng (PG)] as an adulterant for American Ginseng [P. quinquefolius (PQ)] using spectral fingerprints obtained by flow injection mass spectrometry (FIMS). Samples of 100% PQ and 100% PG were physically mixed to provide 90, 80, and 50% PQ. The multivariate FIMS fingerprint data were analyzed using soft independent modeling of class analogy (SIMCA) based on 100% PQ. The Q statistic, a measure of the degree of non-fit of the test samples with the calibration model, was used as the analytical parameter. FIMS was able to discriminate between 100% PQ and 100% PG, and between 100% PQ and 90, 80, and 50% PQ. The probability of identification (POI) curve was estimated based on the SD of 90% PQ. A digital model of adulteration, obtained by mathematically summing the experimentally acquired spectra of 100% PQ and 100% PG in the desired ratios, agreed well with the physical data and provided an easy and more accurate method for constructing the POI curve. Two chemometric modeling methods, SIMCA and fuzzy optimal associative memories, and two classification methods, partial least squares-discriminant analysis and fuzzy rule-building expert systems, were applied to the data. The modeling methods correctly identified the adulterated samples; the classification methods did not.
美国官方分析化学家协会（AOAC INTERNATIONAL）的植物鉴定方法验证指南被应用于检测亚洲人参[Panax ginseng (PG)]作为西洋参[P. quinquefolius (PQ)]掺假物的情况，采用流动注射质谱法（FIMS）获得的光谱指纹图谱进行检测。将100% PQ和100% PG的样品进行物理混合，得到90%、80%和50% PQ的样品。基于100% PQ，使用类软独立建模（SIMCA）对多元FIMS指纹数据进行分析。Q统计量作为测试样品与校准模型不拟合程度的度量，被用作分析参数。FIMS能够区分100% PQ和100% PG，以及100% PQ与90%、80%和50% PQ。基于90% PQ的标准差估计鉴定概率（POI）曲线。通过将100% PQ和100% PG的实验获得光谱按所需比例进行数学求和得到的掺假数字模型，与物理数据吻合良好，并提供了一种简单且更准确的构建POI曲线的方法。两种化学计量学建模方法，即SIMCA和模糊最优联想记忆，以及两种分类方法，即偏最小二乘判别分析和模糊规则构建专家系统，被应用于这些数据。建模方法正确识别了掺假样品；分类方法则未能做到。
