论文部分内容阅读
目的:建立一种基于电子舌对药物的苦度进行预测的方法。方法:以盐酸小檗碱为参比,苦参碱和氧化苦参碱为模型药物,基于25位口尝评价员的口感评价结果和TS-5000Z电子舌传感器的味觉信息数据建立相应的苦度预测模型(BMP),并使用交互验证和残差分析法对模型拟合精度和优度进行评价,对电子舌预测苦味化合物苦度能力进行探索和评价。结果:本研究建立的电子舌对苦参碱和氧化苦参碱的苦度预测模型决定系数R~2分别为0.895 5(P<0.01,n=6)和0.979 3(P<0.01,n=6),均方根误差RMSE分别是0.563 1和0.290 3;交互验证的预测值与真实值之间的相关系数R分别为0.963 9(P<0.01,n=4)和0.953 5(P<0.01,n=4),交互验证均方根误差RMSECV分别为0.306 9,0.276 5;标准化残差在±2.776范围内呈随机分布,显示回归结果较好。结论:本研究建立的模型拟合精度和拟合优度均较高,能够较准确的预测苦参碱和氧化苦参碱的苦度,可以作为苦参碱和氧化苦参碱溶液苦度预测的模型,并为其他药物苦度预测模型的建立提供参考。
Objective: To establish a method based on electronic tongue to predict the bitterness of drugs. Methods: Berberine hydrochloride was used as a reference substance. Matrine and oxymatrine were used as model drugs to establish the corresponding bitterness based on the taste evaluation results of 25 mouth taste assessors and the taste information data of TS-5000Z electronic tongue sensor Prediction models (BMPs). The accuracy and goodness of fit of the models were evaluated by using cross-validation and residual analysis, and the biting ability of the electronic tongue was predicted and evaluated. Results: The determination coefficients of malaria and oxymatrine in electronic tongue established by this study were 0.895 5 (P <0.01, n = 6) and 0.979 3 (P <0.01, n = 6) and the root mean square error (RMSE) were 0.563 1 and 0.290 3, respectively. The correlation coefficients R between the predicted and the true values of cross validation were 0.963 9 (P <0.01, n = 4) and 0.953 5 , n = 4). RMSECV of cross validation was 0.306 9 and 0.2276 5, respectively. The standardized residuals were randomly distributed within the range of ± 2.776, which showed that the regression results were good. Conclusion: The model fitting accuracy and goodness of fit established in this study are both high, which can predict the bitterness of matrine and oxymatrine more accurately and can be used to predict the bitterness of matrine and oxymatrine solutions And provide a reference for the establishment of prediction model of bitterness of other drugs.