基于稳定同位素与矿物元素的蜂蜜产地溯源判别模型的构建

Construction of honey origin traceability discriminant model based on stable isotopes and mineral elements

目的对不同产地蜂蜜的稳定同位素、矿物元素进行含量差异分析,结合多元统计学技术,筛选出有效的产地溯源指标,构建蜂蜜的产地溯源判别模型。方法通过采集甘肃、江苏、安徽、湖南、广西、广东和辽宁7个地区的蜂蜜样本,利用稳定同位素比率质谱法(isotope ratio mass spectrometry,IRMS)和电感耦合等离子体质谱法(inductively coupled plasma mass spectrometry,ICP-MS)测定稳定同位素δ^(13)C、δ^(15)N和22种矿物元素。结合单因素方差分析(analysis of variance,ANOVA)、主成分分析(principle components analysis,PCA)、正交偏最小二乘判别分析(orthogonal partial least squares-discriminant analysis,OPLS-DA)、线性判别分析(linear discriminant analysis,LDA)等多元统计分析方法探讨不同产地蜂蜜判别的可行性。结果不同产地的稳定同位素δ^(13)C、δ^(15)N和22种矿物元素均存在一定的差异。PCA提取的7个主成分方差总贡献率为83.254%,并能够实现不同产地蜂蜜的初步分类。基于OPLS-DA建立的判别模型,能够较好地进行7个地区蜂蜜的产地区分,同时筛选出变量重要性投影(variable importance for the projection,VIP)值大于1的特征差异指标(Cr、As、K、Al、Ti、Cu、δ^(13)C、Li、δ^(15)N、Na、Zn、Be、Mn)并用于Fisher线性判别分析,构建的判别模型判别正确率为100.0%,交叉验证率为95.7%,基本实现了蜂蜜的产地判别。结论本研究表明利用稳定同位素和矿物元素并结合多元统计学分析能够建立蜂蜜的产地溯源模型。

Objective To establish a honey origin traceability discriminant model by analyzing the difference of stable isotopes and mineral elements in honey from different places,combined with multivariate statistical analysis technology and selecting effective origin traceability indicators.Methods Isotope ratio mass spectrometry(IRMS)and inductively coupled plasma mass spectrometry(ICP-MS)were used to determine stable isotopes δ^(13)C,δ^(15)N and 22 kinds of mineral elements by collecting honey samples from 7 regions of Gansu,Jiangsu,Anhui,Hunan,Guangxi,Guangdong and Liaoning.Combined with one-way analysis of variance(ANOVA),principle components analysis(PCA),orthogonal partial least squares-discriminant analysis(OPLS-DA),linear discrimination analysis(LDA)and other multivariate statistical analysis methods were used to explore the feasibility of honey differentiation in different origin.Results The stable isotopes δ^(13)C,δ^(15)N and 22 kinds of mineral elements were different from each other.The total contribution rate of 7 principal components extracted by PCA was 83.254%,and it could realize the preliminary classification of honey from different origin.The discriminative model based on OPLS-DA could well distinguish the origin of honey in 7 regions.At the same time,the characteristic difference indexes with variable importance for the projection value greater than 1(Cr,As,K,Al,Ti,Cu,δ^(13)C,Li,δ^(15)N,Na,Zn,Be,Mn)were screened and used in Fisher discriminant analysis.The accuracy of discriminant model constructed was 100.0% and the cross-validation rate was 95.7%,which basically realized the origin discrimination of honey.Conclusions This study shows that the honey origin traceability discriminant model can be established by using stable isotopes and mineral elements combined with multivariate statistical analysis.