Prediction of minerals, fatty acid composition and cholesterol content of commercial cheeses by near infrared transmittance spectroscopy

Prediction models for the mineral, fatty acid (FA) and cholesterol contents of commercial European cheeses using near infrared transmittance spectroscopy were developed. Cheese samples (n = 145) were from different dairy species and ripening time. Sample spectra were matched with mineral, FA and cholesterol reference data to develop prediction models. Modified partial least squares regressions were validated through cross-validation procedure on the complete dataset (n = 145) and through external validation after dividing the data into calibration (74%) and external validation (26%) sets. Satisfactory models were developed for Ca, P, S, Mg and Zn, and for FA groups (saturated, unsaturated, monounsaturated and polyunsaturated FAs), major FAs (myristic, palmitic and oleic acids) and some minor FAs, whereas cholesterol content could not be predicted with adequate accuracy. Results of the present study are a precursor to at-line utilisation of prediction models for the most abundant cheese minerals and FAs at an industry level.