Prediction of the geographical origin of butters by partial least square discriminant analysis (PLS-DA) applied to infrared spectroscopy (FTIR) data

• Rapid method to accurately predict geographical origin of butter was developed.
• Butters from Morocco were classified by chemometric elaboration of FTIR data.
• IR range 3000–600 cm−1 used bands for lipid, protein components specific to butter.
• PLS-DA model proved 100% accurate for classifying butter samples.
• Discrimination between butters from different regions was fast, simple and accurate.

This study examined the potential of Fourier transform infrared spectroscopy (FTIR) in combination with chemometric methods to discriminate among butters of different regions from Morocco. Chemometric analysis of the data provided by FTIR analysis made it possible to establish links to the food origin of 54 butter samples produced in the Fkih Ben Saleh, Kssiba and Kalaa Sraghna areas. The data of calibration set provided a characteristic pattern, or ‘fingerprint’, relating to the origin of the samples, with good discriminant power. Two models using principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were built. The PCA model was able to describe the studied system by using four principal components with a value of explained variance of 98%. The PLS-DA model accurately classified the butter samples of an external validation subset with prediction ability of 100%. The proposed methods, if compared to other techniques, have the main advantage in allowing very rapid measurements and results characterized by high accuracy and precision.