In this study, a total of 138 milk tea powder samples were collected from the market of Inner Mongolia and classified into two categories based on the manufacturing process and major raw materials: traditional wet-processed and modern dry-mixed powder. In addition, 40 traditional cheese samples and 15 solid milk-containing cheese-like product samples were obtained from Inner Mongolia’s different regions. For comparison, five samples of raw milk, milk casein (not included in the total), milk powder, and tofu were also collected. Eighteen amino acids (AAs) were quantified using the national standard method and subjected to statistical analyses using orthogonal partial least squares discriminant analysis (OPLS-DA) and hierarchical cluster analysis (HCA). Results revealed significant differences in AA contents and fingerprint patterns between the two kinds of milk tea powder, as well as between traditional cheese and solid milk-containing products. OPLS-DA clearly differentiated between wet- and dry-processed milk tea powder, and between traditional cheese and solid milk-containing products. Notably, however, raw milk-derived milk tea powder and traditional cheese were closely clustered with the control samples (raw milk, milk casein, and milk powder). Tofu samples were clustered separately and distinctly from all other groups, indicating no incorporation of soybean-derived ingredients. External validation accuracies for the OPLS-DA models based on the absolute and relative contents of AAs in milk tea powder were 70.8% and 75.4%, respectively, and those for traditional cheese were 75.8% and 95.0%, respectively. These findings confirm the feasibility of authenticating milk tea powder and traditional cheese using AA fingerprints.

This work was undertaken in order to demonstrate the feasibility of establishing a model based on fatty acid (FA) fingerprints to discriminate between beef from pasturing and barn feeding systems. A total of 91 samples of biceps femoris, longissimus dorsi, and costal subcutaneous fat from pastured and barn fed cattle from eight counties/banners in the dominant beef cattle farming belt of Inner Mongolia were collected for determination of fatty acids (FAs) by gas chromatography (GC), and principal component analysis (PCA) and descriptive statistics were conducted on the obtained data. Furthermore, soft independent modeling of class analogy (SIMCA) was used to establish a model to authenticate pastured and barn-fed beef. Results indicated that all samples were clustered into pasturing and barn feeding groups, that beef samples from grassland and agricultural areas were also clustered separately, that grassland beef samples from Hulun Buir and Xilingol could be separated, that barn-fed beef samples from Horqin Left Wing Rear Banner and Zhenglan Banner in the eastern region were grouped separately, and that barn-fed beef samples from Helingeer county and Uxin Banner in the mid-western regions tended to be separated. The contents of n-3 polyunsaturated fatty acids (PUFAs), α-C_{18:3 n3} (α-linolenic acid), C_18:0 and C_14:0 in pastured beef were significantly higher than those in barn-feeding beef, and the contents of n-3 PUFAs and α-linolenic acid increased by 1.6 and 2.2 times, respectively. On the other hand, the contents of n-6 PUFAs, C_{18:1 n9c} and C_{18:2 n6c} (linoleic acid) in barn-fed beef were significantly higher than those in pastured beef. The ratio of n-6 to n-3 PUFAs in pastured beef was 3.6, closer to the ideal value. However, the paired t-test and blocked F test on means could not distinguish feeding systems or regions in terms of FA profiles, implying that the conventional statistics has limited ability to evaluate the overall pattern of indicator datasets. PCA performed on FA profiles of the three cuts showed better clustering effect on pastured and barn-fee beef than all samples. FAs of costal subcutaneous fat exhibited the farthest clustering distance for both pastured and barn-fee beef. The internal and external verification accuracy of the established SIMCA model were 100% and 92.6%, respectively. FA-based modeling is feasible to distinguish pastured from barn-fed beef. This study can provide innovative ideas and methods for further studies on meat authentication.