Prediction of the protein and lipid content of Coalho cheese using spectroscopy

Abstract

Coalho cheese is a traditional dairy product consumed in Brazil. Official methodologies for determining the cheese composition have disadvantages related to costs and analysis time, but Near Infrared Region Spectroscopy (NIRS), a faster methodology, is an alternative that can be used to obtain fast and reliable data. The aim of this study was to use NIRS associated with multivariate analysis to predict the content of proteins and lipids in Coalho cheese. 80 samples of the product were submitted to the official analysis methods and NIRS. The spectra were obtained in the region 10000 - 4000 cm-1, in reflectance mode. The multivariate calibration method used was partial least squares regression (PLS) with PLS interval algorithm. It was found that the mean square error of prediction for the chosen model for lipids was 1.39% and, for proteins, 2.73%. The NIRS showed imprecision for protein prediction in Coalho cheese, but, for the lipid content, the NIRS-PLS methodology showed satisfactory prediction results, which shows its potential for the Dairy Industries.