目的 建立楮实子Broussonetiae Fructus中槲皮素、木犀草素、芹菜素、大黄素、大黄素甲醚、鹅掌楸碱、白屈菜红碱、氯化两面针碱、岩藻甾醇、胡萝卜苷、β-谷甾醇11种成分定量检测方法，同时检测其醇溶性浸出物、总灰分和酸不溶性灰分含量，并结合化学模式识别和Logistic回归分析对不同产地楮实子进行等级预测。方法 采用Synergi Max-RP色谱柱；以乙腈-0.3%磷酸溶液为流动相，梯度洗脱；检测波长360 nm（槲皮素、木犀草素和芹菜素）、271 nm（大黄素、大黄素甲醚、鹅掌楸碱、白屈菜红碱和氯化两面针碱）和210 nm（岩藻甾醇、胡萝卜苷和β-谷甾醇）；采用外标法检测楮实子中11种成分含量。按《中国药典》2020年版四部检测楮实子中醇溶性浸出物、总灰分和酸不溶性灰分含量。利用SPSS 26.0和SIMCA 14.1软件对14个定量检测指标结果进行化学模式识别分析，采用Logistic回归分析建立不同产地楮实子的等级预测模型，并进行验证。结果 11种成分在各自质量浓度范围内线性关系良好（r＞0.999）；平均加样回收率为96.98%～100.08%，RSD为0.71%～1.84%；精密度、重复性和稳定性良好（RSD均＜2.0%）。45批楮实子中槲皮素、木犀草素、芹菜素、大黄素、大黄素甲醚、鹅掌楸碱、白屈菜红碱、氯化两面针碱、岩藻甾醇、胡萝卜苷和β-谷甾醇质量分数分别为（2.319±0.377）、（1.957±0.342）、（4.818±0.779）、（0.301±0.054）、（0.701±0.158）、（0.936±0.158）、（1.771±0.295）、（0.431±0.085）、（0.123±0.037）、（0.088±0.023）、（0.409±0.084）mg/g；醇溶性浸出物、总灰分和酸不溶性灰分含量分别为（18.1±2.6）%、（6.8±0.6）%和（0.9±0.4）%。主成分分析结果显示有2个主成分的特征值大于1，45批楮实子聚为3类；因子分析显示S17～S31排序靠前、S1～S16排序靠中、S32～S45排序靠后；正交偏最小二乘判别分析显示芹菜素、大黄素甲醚、槲皮素、木犀草素、白屈菜红碱和β-谷甾醇是不同产地楮实子质量差异因子。Logistic回归分析显示45批楮实子所对应的预测归属等级明确，拟合概率P值均大于98.0%。结论 HPLC多成分定量检测、化学模式识别及Logistic回归分析模型操作便捷、结果准确，可用于不同产地楮实子的等级预测，为中药楮实子的质量评价标准制定提供数据参考。

Objective To establish a quantitative detection method for 11 components of quercetin, luteolin, apigenin, emodin, physcion, liriodenine, chelerythrine, nitidine chloride, fucosterol, daucosterol and β-sitosterol in Chushizi (Broussonetiae Fructus), at the same time, the contents of alcohol-soluble extract, total ash and acid-insoluble ash were detected, and the grades of Broussonetiae Fructus from different producing areas were predicted by chemical pattern recognition and Logistic regression analysis. Methods All samples were analyzed by Synergi Max-RP column and eluted with acetonitrile-0.3%phosphoric acid performing gradient elution, and the detection wavelength were 360 nm (quercetin, luteolin and apigenin), 271 nm (emodin, physcion, liriodenine, chelerythrine and nitidine chloride ) and 210 nm ( fucosterol, daucosterol and β-sitosterol). The contents of 11 components were detected by external standard method. The contents of alcohol-soluble extract, total ash and acid-insoluble ash in Broussonetiae Fructus were determined according to Chinese Pharmacopoeia (Volume IV). Using SPSS 26.0 and SIMCA 14.1 software, chemical pattern recognition analysis was performed on the results of 14 quantitative detection indicators. Logistic regression analysis was used to establish and verify the grade prediction model of Broussonetiae Fructus from different producing areas.Results The linear relationship was good within the respective quality concentration ranges for all 11 components (r > 0.999), the average recoveries were 96.98%—100.08% with RSDs of 0.71%—1.84%. The RSDs of precision, repeatability and stability were all less than 2.0%. The mass fractions of quercetin, luteolin, apigenin, emodin, physcion, liriodendrine, chelerythrine, nitidine chloride, fucosterol, daucosterol and β-sitosterol in 45 batches of Broussonetiae Fructus were (2.319 ±0.377), (1.957 ±0.342), (4.818 ±0.779), (0.301 ±0.054), (0.701 ±0.158), (0.936 ±0.158), (1.771 ±0.295), (0.431 ±0.085), (0.123 ±0.037), (0.088 ±0.023) and (0.409 ±0.084) mg/g, respectively. The contents of alcohol-soluble extract, total ash and acid-insoluble ash were (18.1 ±2.6)%, (6.8 ±0.6)% and (0.9 ±0.4)%, respectively. Principal component analysis showed that the eigenvalues of two principal components were greater than 1, and 45 batches of BroussonetiaeFructus were clustered into three categories. Factor analysis showed that S17—S31 was in the front, S1—S16 was in the middle, and S32—S45 was in the back. Orthogonal partial least squares discriminant analysis showed that apigenin, physcion, quercetin, luteolin, chelerythrine and β-sitosterol were quality difference factors of Broussonetiae Fructus from different producing areas. Logistic regression analysis showed that the predicted attribution levels corresponding to 45 batches of Broussonetiae Fructus were clear, and the fitting probability P values were all greater than 98.0%. Conclusion HPLC multi-component quantitative detection, chemical pattern recognition and Logistic regression analysis model are convenient and accurate, which can be used to predict the grade of Broussonetiae Fructus from different producing areas, and provide data reference for the establishment of quality evaluation standard of Broussonetiae Fructus.

R286.2

河南省医学科技攻关计划联合共建项目（LHGJ20210992）