Identification and quantification of adulteration in frozen-thawed meat of different breeds by NIR spectroscopy and chemometrics

This work aims to compare the characteristic near-infrared absorption bands of different breeds of meats and to develop a qualitative and semi-quantitative method for the detection of adulteration in various types of frozen meat based on near-infrared spectroscopy and chemometrics methods. The adulterated samples were in various combinations of pork, beef and mutton adulterated with different proportions of pork, beef, lamb, duck and chicken. The results showed that the accuracy of the PLS-DA model that identified all adulteration types together was the lowest (AUCP= 0.8265), while the accuracy of the PLS-DA model of lamb-chicken and beef-chicken was the highest (AUCP= 1). AUCP of other adulteration types was higher than 0.92. And the PLSR model constructed by all the adulteration types together had the worst result (Rp2= 0.5348, RMSECV=0.1895, RPD=1.48). The results of PLSR prediction models constructed by the adulteration types separately all reached the expectation (Rp2= 0.8707~0.9703,RMSECV= 0.0637~0.1089,RPD= 2.78~5.80). The results of the PLS-DA models and PLSR models using characteristic band spectral information are basically the same or slightly improved compared with those results of the models using full-band spectral information. Therefore, NIR spectroscopy and chemometrics methods can directly qualitatively identified all adulteration types, but it is more accurate to semi-quantitatively predict the proportion of adulteration according to different adulteration types. The near-infrared absorption of moisture and fat contributed more to the detection of meat adulteration.