Application Of An Electronic Nose For Measurements Of Goaty Flavor In Goat Milk

Goaty flavor has a significant negative influence on the the quality of goat milkproducts. The analysis of goaty flavor has been commonly carried out by human assessmentand the most important problems of this method are measurement standardization, stability,and reproducibility. Therefore, the aim of this work was to investigate the potential ofelectronic nose as a useful tool to discriminate the intensity of goaty flavor in goats’ milk.First, a method for the simultaneous determination of the free fatty acids (FFA) in goat milksuch as caproic acid, caprylic acid and decanoic acid, which are responsible for goaty flavor,was developed by multilevel solvent extraction-gas chromatography. Extractions were carriedout with chloroform/methanol (2:1, v/v) and sodium carbonate solution. The fatty acid methylesters (FAMEs) which were obtained by the esterification of the FFA with borontrifluoride-methanol solution was qualified and quantified by gas chromatography equippedwith flame ionization detector. Over the concentration in the range of5.11~163.52mg/L forcaproic acid,5.05~161.73mg/L for caprylic acid and7.54~241.38mg/L for decanoic acid, allthe decisive coefficients (R2) were above0.998. The limits of detection (LOD) were0.16mg/L for caproic acid,0.19mg/L for caprylic acid and0.25mg/L for decanoic acid. The limitsof quantification (LOQ) were0.54mg/L for caproic acid,0.63mg/L for caprylic acid and0.83mg/L for decanoic acid. The average recoveries of the three kinds of FFA ranged from88.5%to103.4%with relative standard deviations (RSDs) between1.98%and6.09%. Themethod is economical, accurate, and sensitive, which may be used for determination of theFFAs in milk. Afterwards, an electronic nose (PEN3) based on metal oxide sensors was usedfor goaty flavor measurements of goat milk samples. In order to study the relationshipsbetween e-nose data, sensory data and levels of the free fatty acids (FFA) multivariate partialleast lquare regressions (PLS) was carried out. All the PLS models’ decisive coefficients (R2)were up to0.90. Principal component analysis (PCA) can discriminate goat milk sampleswhich have different intensity goaty flavor. Fisher discriminant analysis (FDA) and BP neuralnetwork were carried out to identify the goaty flavor intensity. The prediction accuracies ofprediction sets were98.2%and100.0%, respectively. These results will be helpful for theapplication of electronic nose to evaluate the goaty flavor in goat milk samples. In addition, the electronic nose (PEN3) was used to discriminate the variety of lactic acid bacteria in goatyogurt samples. The volatile components emanating from goat yogurt samples were gatheredby the electronic nose. The electronic nose technologies for discriminating the variety oflactic acid bacteria in goat yogurt were established through PCA, fisher linear discriminantanalysis (FLDA) and BP neural network. The results showed that both PCA and FLDA candiscriminate the variety of lactic acid bacteria, but FLDA was more effective than PCA. Theprediction correct rate of FLDA and BP neural network were100.0%and98.4%, respectively.These results will be useful for the application of electronic nose to discriminate the variety oflactic acid bacteria in goat yogurt samples.