Previsão do teor de proteínas e lipídios do queijo de coalho utilizando a espectroscopia no infravermelho próximo e calibração multivariada

Abstract

The “coalho” cheese is a delicacy incorporated into the menu of Brazilians. It is a significant food from a territorial identity, constituting a cultural food from Northeast Brazil. The technology employed in cheese production is simple, usually occurs in small dairies. Conventional methods for food analysis takes time, equipment, specific labors and generate wastes. Development of instrumental analysis techniques allowed the creation of rapid, nondestructive technology that allow to estimate the composition of food. Thus, the objectiveof this study was to provide content of proteins and lipids of “coalho” cheese, using spectroscopy in the near infrared region associated with multivariate calibration. Therefore, 80 samples of “coalho” cheese were purchased in local shops to which were performed the analysis by conventional methods (Kjeldahl for protein and Gerber to lipids) and NIR spectroscopy. Spectroscopic analyzes were performed in the near infrared region (10000 - 4000 cm -1) in the diffuse reflectance mode with resolution of 8 cm -1, and average of 16 scans. For a better conditioning of the matrix data, different pre-processing were tested, and the 1st derivative with filter Savitzky - Golay using window 19 points and 2nd degree polynomial of the most effective.The calibration method was partial least squares regression (PLS) with variable selection variables using the algorithm interval PLS (iPLS). Overall, we found that the mean square error of prediction - RMSEP for the chosen model for the lipid content was 1.39%. Regarding the variable proteins a RMSEP value of 2.73% was obtained. The near infrared spectroscopy showed imprecision for protein prediction, although for other cheese varieties, the methodology has been successfully applied. For the lipid content the determination of fat in samples with complex matrices such as “coalho” cheese showed the advantages of association between NIR-PLS.