基于吉祥草高效液相色谱（HPLC）指纹图谱，结合化学模式识别评价20批吉祥草（S1~S20）的质量，并测定其中芦丁、人参皂苷Rb1、薯蓣皂苷、薯蓣皂苷元等4种化学成分的含量。吉祥草样品经75%（体积分数）乙醇溶液回流提取2次，过滤、合并滤液、旋转蒸发至干，得到的残渣用甲醇溶解并定容至10 mL，过0.45 μm微孔滤膜，滤液供HPLC分析。采用中药色谱指纹图谱相似度评价系统（2012A版）建立20批吉祥草HPLC指纹图谱，并对其进行相似度评价。结合化学模式识别，进行聚类分析、主成分分析和偏最小二乘判别分析，并筛选差异性成分。结果表明：从20批吉祥草HPLC指纹图谱中，选取了13个共有峰，指认其中4个成分，分别为芦丁（峰5）、人参皂苷Rb1（峰11）、薯蓣皂苷（峰12）、薯蓣皂苷元（峰13）；20批吉祥草图谱与对照图谱相似度为0.546~0.942；经聚类分析和偏最小二乘判别分析，20批吉祥草均被分成3类，其中S7、S10、S12、S16、S18、S17、S15、S3、S4、S1、S13、S2、S14、S19、S6、S8、S20、S9为第1类，S11为第2类，S5为第3类；经主成分分析，主成分1~4的累积方差贡献率为85.374%，20批吉祥草中S5的综合得分较高，其次是S11；采用变量重要性投影法筛选出7个差异性成分，分别为峰13（薯蓣皂苷元）、峰2、峰7、峰5（芦丁）、峰12（薯蓣皂苷）、峰1和峰6；芦丁、人参皂苷Rb1、薯蓣皂苷、薯蓣皂苷元的质量在一定范围内与其对应的峰面积呈线性关系，检出限（3S/N）为0.005~0.020 mg·g^-1；对S2进行加标回收试验，4种化学成分的回收率为97.4%~102%，测定值的相对标准偏差（n=6）为1.1%~2.6%；方法用于20批吉祥草分析，芦丁、人参皂苷Rb1、薯蓣皂苷和薯蓣皂苷元的测定值分别为0.249~0.984 mg·g^-1，0.431~5.851 mg·g^-1，0.007~0.261 mg·g^-1和0.003~0.095 mg·g^-1。

Based on the HPLC fingerprint of Reineckia carnea, the quality of 20 batches of Reineckia carnea (S1-S20) was evaluated with chemical pattern recognition, and 4 chemical components, including rutin, ginsenoside Rb1, dioscin and diosgenin, were determined. The sample of Reineckia carnea was extracted twice by reflux with 75% (φ) ethanol solution and filtered, and the combined filtrate was evaporated by rotation to dryness. The residue obtained was dissolved and made its volume up to 10 mL with methanol. The solution was passed through 0.45 μm microporous membrane, and the filtrate was used for HPLC analysis. The HPLC fingerprints of 20 batches of Reineckia carnea were established and their similarity was evaluated with the similarity evaluation system for chromatographic fingerprint of traditional Chinese medicine (2012A version). Cluster analysis, principal component analysis and partial least squares-discriminant analysis were performed combined with chemical pattern recognition, and the differential components were screened. The results showed that 13 common peaks were selected from the HPLC fingerprints of 20 batches of Reineckia carnea, and 4 components were identified from them, including rutin (peak 5), ginsenoside Rb1 (peak 11), dioscin (peak 12) and diosgenin (peak 13). The similarity between the fingerprints of 20 batches of Reineckia carnea and the control profile was in the range of 0.546-0.942. Cluster analysis and partial least squares-discriminant analysis showed that 20 batches of Reineckia carnea were divided into 3 classes, and S7, S10, S12, S16, S18, S17, S15, S3, S4, S1, S13, S2, S14, S19, S6, S8, S20 and S9 were class 1, S11 was class 2, and S5 was class 3. Principal component analysis showed that the cumulative variance contribution rate of principal components 1-4 was 85.374%, and S5 had the highest comprehensive score among 20 batches of Reineckia carnea, followed by S11. Variable importance projection method was used to screen out 7 differential components, which were peak 13 (diosgenin), peak 2, peak 7, peak 5 (rutin), peak 12 (dioscin), peak 1 and peak 6. Linear relationships between the masses of rutin, ginsenoside Rb1, dioscin and diosgenin and their corresponding peak areas were kept in definite ranges, with detection limits (3S/N) in the range of 0.005-0.020 mg·g^-1. Test for recovery was made by standard addition method on S2, giving results in the range of 97.4%-102%, and RSD (n=6) of the determined values were ranged from 1.1% to 2.6%. This method was applied to the analysis of 20 batches of Reineckia carnea, and the determined values of rutin, ginsenoside Rb1, dioscin and diosgenin were 0.249-0.984 mg·g^-1, 0.431-5.851 mg·g^-1, 0.007-0.261 mg·g^-1 and 0.003-0.095 mg·g^-1, respectively.