Rapid classification and quantification of mixture containing Sumbawa’s horse milk and soybean milk using Near Infrared spectroscopy combined with chemometrics
Horse milk (HM) can be described as a distinctive dairy product with increased nutritional content. Soybean milk (SM) is widely mixed with animal milk due to its lower price than other milk. The adulteration of HM with SM could pose economic fraud. This study aims to develop a rapid and reliable approach for detecting HM adulteration with SM using Near Infrared (NIR) spectroscopy combined with chemometrics. The binary mixtures of HM and SM were prepared to develop calibration and validation models. The classification (with PCA and PLS-DA) and quantification analysis (using PLS regression) were applied to interpret NIR spectral data. The PCA model assessed seven principal components and no outlier data based on Summary Fit and Hotelling’s T2 Range column plot. PCA and PLS-DA were utilized to distinguish between pure HM and adulterated SM in score plot and biplot. The VIP scores were generated at wavelengths 1908-1952, 2554, and 2557-2592 nm using the PLS-DA model. A combination of OH-bound, CH-bound, and NH functional groups was categorized as the wavenumber selected in VIP scores. The regression result was obtained with y = x-0.00000001026 with R2 = 0.9985. The NIRPLSR model could successfully distinguish HM from SM adulteration with an accuracy of R2 > 0.999 and Q2 > 0.80 and precision of RMSEE = 0.0164 and RMSECV = 0.2556 with good prediction. A dairy product authenticity study could be performed using NIR spectra combined with chemometric techniques.