Prediction of the Fatty Acid Profiles of Iberian Pig Products by Near Infrared Spectroscopy: A Comparison Between Multiple Regression Tools and Artificial Neural Networks

Abstract

In this study, the feasibility of predicting the lipid profiles of Iberian ham and shoulder samples by using near infrared (NIR) spectroscopy was evaluated. Gas chromatography analysis was the reference method used. The muscles analyzed and recorded by NIR spectroscopy were 76 Biceps femoris for Iberian hams and 72 Brachiocephalicus for Iberian shoulders. NIR calibrations were carried out by using two methods: modified partial least squares regression (MPLS) and artificial neural networks (ANN). With the MPLS method, it was possible to obtain equations with regression’s coefficients (RSQ) of > 0.5 for 5 individual fatty acids and 3 summations: polyunsaturated fatty acids, n3 and n6. The use of neural networks made it possible to find equations with RSQ of > 0.5 for 10 individual fatty acids, all of which are present in over 90% of the samples, and 5 summations of saturated, monounsaturated, and polyunsaturated fatty acids (SFA, MUFA, PUFA), n3 and n6, finding that the calibration curves of the fatty acids C18:1, C18:2n6, and C18:3n3 presented RSQ’s of > 0.7. The results obtained indicate that NIR spectroscopy could be a very useful technology for the quality control of cured products as it allows estimating the main fatty constituents quickly and without using reagents.