Feasibility study on using visible-near infrared spectroscopy coupled with factorial discriminant analysis technique to identify sheep milk from different genotypes and feeding systems. An Abstract from Journal of Near Infrared Spectroscopy

A.M. Mouazen,^a,* S. Dridi,^b H. Rouissi,^c J. De Baerdemaeker^d and H. Ramon^d
^aNatural Resources Department, Cranfield University, MK43 0AL Cranfield, United Kingdom. E-mail: a.mouazen@cranfield.ac.uk
^bLaboratory of Physiology, Immunology, and Genetics of Domestic Animals, Kasteelpark Arenberg 30, B-3001 Heverlee, Belgium
^cDepartment of Animal Production, Higher Institute of Agriculture of Mateur, route de Tabarka 7030, Mateur, Tunisia
^dDivision of Mechatronics, Biostatistics and Sensors (MeBioS), Faculty of Bioscience Engineering, Kasteelpark Arenberg 30, B-3001 Heverlee, Belgium

Quick detection of differences in raw milk samples due to feeding and genotype is needed for quality control of milk and milk products. This study was undertaken to evaluate the potential of visible (vis) and near infrared (NIR) spectroscopy to discriminate between differences in sheep milk that are attributed to different feedings and genotypes. A mobile, fibre-type, vis-NIR spectrophotometer (Zeiss Corona 1.7 visnir fibre, Jena, Germany), with a light reflectance measurement range of 306.5–1710.9 nm was used to measure the light reflectance from 50 raw milk samples obtained from two experimental trials in Tunisia. Out of these 50 samples, 30 samples were from the first trial and were equally obtained from two different genotypes and 20 samples were from the second trial and were equally obtained from two different feedings of the same genotype. Milk samples were classified into four different classes, namely, Comisana genotype with pasture feeding (CO), Sicilo–Sarde genotype with pasture feeding (SS), Sicilo–Sarde genotype feeding on Scotch horse bean meal (SSB) and Sicilo–Sarde genotype feeding on soybean meal (SSS). Another milk sample classification based on three classes for three different feedings of one genotype (Sicilo–Sarde) was considered by removing CO. The factorial discriminant analysis (FDA) was applied on the first five principal components (PCs) obtained from the principal component analysis (PCA) performed on the vis-NIR spectra in order to classify milk samples into four and three assigned groups. Based on four-group classification, correct classification (CC) of 92.5% and 95% was observed for the calibration and validation data sets, respectively. For three group classification, CC was 100% and 88.9% for the calibration and validation data sets, respectively. Nevertheless the number of samples of the second trail (different feedings) was smaller (20 samples) compared to that of the first trial (30 samples), the FDA provided better discrimination of differences attributed to feeding compared to that attributed to different genotype. It was concluded that vis-NIR spectroscopy coupled with the FDA classification technique is a powerful tool to discriminate between genotypes and feeding differences that exist in raw sheep milk samples.