基于连续分类策略的饲用鱼油掺假陆生动物油脂红外光谱鉴别分析

A Sequential Classification Strategy Applied to the Detection of Terrestrial Animal Lipid in Fish Oil by MID-Infrared Spectroscopy

与其他动物油脂相比,饲用鱼油的营养价值高、产量低、提炼工艺复杂,其真实性的鉴定有利于市场的正常运行和消费者权益的保障。本研究提出一种基于红外光谱的连续分类策略,并将其应用于饲用鱼油中违法掺假陆生动物油脂的鉴别分析。实验收集动物油脂样品共50个(鱼油12个、猪脂10个、鸡油9个、牛脂10个、羊脂9个),采用均匀混合法制备饲用鱼油中掺加陆生动物油脂的样品160个。采用主成分分析(PCA)方法进行用饲用鱼油中掺假陆生动物油脂红外光谱鉴别分析的可行性分析。结果表明:纯鱼油和掺假陆生动物油脂鱼油之间得到了较好的区分;其他掺假陆生动物油脂鱼油之间有一定的鉴别分析潜力。基于偏最小二乘判别分析(PLS-DA)和单类别偏最小二乘法(OC-PLS),第一步,建立检测鱼油真实性的单类别筛查模型;第二步,建立多类别陆生动物油脂掺假的鉴别模型,探讨了两种陆生动物油脂类别划分方式对鉴别模型的影响。研究表明:单类别筛查模型成功区分了纯鱼油和掺假鱼油,识别率和拒绝率均为100%,误判率为0%;按照纯鱼油、猪脂掺假鱼油、鸡油掺假鱼油、牛脂掺假鱼油和羊脂掺假鱼油进行分类,多类别鉴别模型的识别率和拒绝率均大于80%,误判率均在15%以下;按照纯鱼油、非反刍动物油脂掺假和反刍动物油脂掺假进行分类,多类别鉴别模型的识别率和拒绝率均提升至90%以上,误判率减小至7%以下。在提出的连续分类策略中,中红外光谱技术结合化学计量学可以用于高效筛查鱼油中是否掺假陆生动物油脂,并且可以进一步确证掺假陆生动物油脂种属源。该方法作为一种快速筛查与确证分析工具可满足大样本量鱼油中陆生动物油脂掺假的检测需求。

Compared with other animal fat,fish oil has the traits of the high demand of extraction technology,low oil yield and high nutritive value.Authenticity appraisal of fish oil is conducive to the normal operation of the feed market and the protection of consumer rights and interests.In the present study,a Sequential Classification Strategy(SCS)was proposed and applied to identify illegally adulterated terrestrial animal lipid in fish oil.A total of 50 animal fat samples(12 fish oil,10 lard,9 chicken oil,10 tallow and 9 suet)were collected in this experiment,and 160 adulterated fish oil samples with terrestrial animal fat were prepared by homogeneous mixing method.Exploratory research based on the principal component analysis(PCA)method was used to identify the feasibility of identification analysis.The results showed that pure fish oil and adulterated fish oil were well differentiated.The species identification of terrestrial animal fat adulterants is potential.Based on partial least squares-discriminant analysis(PLS-DA)and one class-partial least squares(OC-PLS),the first step was to establish a one class screening model to detect the authenticity of fish oil.In the second step,the identification model of multi-class adulterations(two types of classification)of terrestrial animal lipid was established.Results show that the one-class screening model distinguished pure fish oil from adulterated fish oil,and the recognition rate and rejection rate of the one-class screening model were both 100%,the first multi-class model for the classification of adulterants(lard,chicken oil,suet,tallow)in pure fish oil performed well with the recognition rate and rejection rate higher than 80%and the mis-discrimination ratio lower than 15%,the second multi-class model for the classification of adulterants(ruminant animal lipid,non-ruminant animal lipid)performed better than the first multi-class model with the recognition rate and rejection rate higher than 90%and the mis-discrimination ratio lower than 7%.With the proposed SCS,infrared spectroscopy combined with chemometrics can be used to identify pure fish oil form adulterated fish oil.Furthermore,the species source of adulterants can be recognized effectively.Finally,we can suggest this type of application as a potential tool to assist the feed industry and regulatory organisms in food quality control,allowing detection in feeding fish oil through direct,fast and reliable screening analyses.