目的 采用电子舌技术研究羟丙基-β-环糊精（HP-β-CD）质量浓度（C）变化对苦味化合物及苦味中药的抑苦规律。方法 以盐酸小檗碱、氧化苦参碱、苦参水煎液、穿心莲水煎液为苦味载体，基于口尝评价结果（△I）和电子舌信息（△I_e），分别建立△I-C、△I_e-C 2个抑苦规律模型、探索△I-△I_e两者的抑苦效果预测模型，并使用交叉验证和残差分析法对预测模型的拟合精度和优度进行评价。结果 对4种苦味载体均建立了良好的△I-C威布尔抑苦规律模型，决定系数（R²）依次为0.999 6、0.987 9、0.996 4、0.998 4（P<0.01）；盐酸小檗碱、苦参水煎液和穿心莲水煎液的6个（每个载体2根传感器）△I_e-C威布尔抑苦规律模型的R²依次为0.996 5、0.991 6、0.997 3、0.989 3、0.999 6、0.999 1（P<0.01）；相应的6个△I-△I_e线性抑苦效果预测模型的R²依次为0.989 1、0.968 3、0.989 0、0.982 0、0.977 9、0.986 1（P<0.01）；上述6个预测模型交叉验证的相关系数（R）依次为0.986 0、0.997 3、0.988 4、0.960 8、0.980 2、0.983 9（P<0.01）。测试质量浓度范围内的氧化苦参碱对电子舌4根传感器均无随浓度变化的差异性响应，因此未能建立各类模型。结论 基于电子舌方法得到了随HP-β-CD质量浓度变化的抑苦规律，建立了以HP-β-CD质量浓度或电子舌数据为基础的预测模型，可用于相关抑苦效果预测。部分苦味化合物对电子舌的响应没有相关规律，有待电子舌技术的进一步研发。

Objective To study the bitterness inhibition law of hydroxypropyl-β-cyclodextrin (HP-β-CD) concentration (C) on the bitter compounds and bitter Chinese herbal medicine, and to explore the application of electronic tongue in this study. Methods Berberine, oxymatrine, Sophora flavescens, and Andrographis paniculata decoction were used as bitterness vectors to establish two models of bitterness inhibition law about △I-C and △I_e-C, and to explore the prediction model of bitterness inhibition effect about △I-△I_e, based on the oral taste evaluation results (△I) and electronic tongue information (△I_e). Then, fitting precision and fitting goodness of the prediction model were evaluated with cross-validation and residual analysis. Results In this study, good Weibull model of bitterness inhibition pattern about △I-C were established for all the four bitterness vectors, the decision coefficient R² are as followed:0.999 6, 0.987 9, 0.996 4, and 0.998 4 (P<0.01); The decision coefficient R² of six (two sensors per vector) models of bitterness inhibition law about △I_e-C of berberine, S. flavescens, and A. paniculata decoctions were as followed:0.996 5, 0.991 6, 0.997 3, 0.989 3, 0.999 6, and 0.999 1 (P<0.01); The decision coefficient R² of six corresponding linear prediction models of bitterness inhibition effect about △I-△I_e were as followed:0.989 1, 0.968 3, 0.989 0, 0.982 0, 0.977 9, and 0.986 1 (P<0.01); The correlation coefficient R calculated by correlation coefficient of six prediction models above were as followed:0.986 0, 0.997 3, 0.988 4, 0.960 8, 0.980 2, and 0.983 9 (P<0.01); No model was established for oxymatrine within the range of tested concentration in this research, as it didn't respond to the four sensors with varied concentration. Conclusion Based on this method, the bitterness inhibition law of HP-β-CD with changed concentration was obtained. Prediction models based on HP-β-CD concentration or electronic tongue data were also established, they can be used to predict the relative bitterness inhibition effect. Part of the bitter compounds didn't response to the electronic tongue regularly, remain further research and development of electronic tongue technology.

国家自然科学基金青年基金项目（81001646）；国家自然科学基金面上项目（81774452）；河南中医学院省属高校基本科研业务费优青培育项目（2014KYYWF-YQ01）；河南省中医管理局中医药科学研究专项课题（2014ZY02066）