Prediction of fatty acid composition of fresh and freeze-dried cheeses by visible-near-infrared reflectance spectroscopy

Visible-near-infrared reflectance spectroscopy was used to predict the fatty acid (FA) composition of 445 cows' and goats' fresh and freeze-dried cheeses. The FA composition of samples was analysed by reference method. Samples were scanned (400-2500 nm) and predictive equations were developed using modified partial least squares with both cross-validation and external validation. The coefficient of determination in external validation ((RV)-V-2) and residual predictive deviation in the validation set (RPDV) were good for total saturated (0.89, 2.84), monounsaturated (0.90, 2.90), polyunsaturated (0.88, 2.54), trans (0.92, 3.24) FA, butyric (0.84, 2.43), capric (0.92, 2.96), myristic (0.89, 2.94), palmitic (0.91, 3.32), vaccenic (0.88, 2.63), and rumenic (0.90, 2.70) acids in fresh cheeses. Approximate or poor predictions were obtained for caproic, caprylic, lauric, stearic, oleic, linoleic, and linolenic acids ((RV)-V-2 <= 0.80; RPDV <= 2.08). The quantification was significantly more accurate for caprilic, capric, and linoleic acids with freeze-dried cheeses compared with fresh cheeses. (C) 2008 Elsevier Ltd. All rights reserved.