AUTHOR=Molle Giovanni , Cabiddu Andrea , Decandia Mauro , Sitzia Maria , Ibba Ignazio , Giovanetti Valeria , Scanu Giuseppe , Addis Margherita , Caredda Marco TITLE=Can FT-Mid-Infrared Spectroscopy of Milk Samples Discriminate Different Dietary Regimens of Sheep Grazing With Restricted Access Time? JOURNAL=Frontiers in Veterinary Science VOLUME=Volume 8 - 2021 YEAR=2021 URL=https://www.frontiersin.org/journals/veterinary-science/articles/10.3389/fvets.2021.623823 DOI=10.3389/fvets.2021.623823 ISSN=2297-1769 ABSTRACT=Milks obtained from ruminants grazing natural pastures and some forage crops may worth a plus value as compared to milks obtained to stall fed ruminants, due to their putative higher content of beneficial fatty acids (FA). Fourier Transformed Mid-infrared (FT-MIR) analysis of FA can help to distinguish milks coming from different areas and, possibly, different feeding systems. The objective was to discriminate sheep milks sourced from dairy sheep rotationally grazing for 2, 4 or 6 hr/d Italian ryegrass or berseem clover. To test this hypothesis a data-mining study was undertaken using a database of 1230 individual milk spectra. Data were elaborated by Principal Component Analysis (PCA) and analyzed by Linear Discriminant analysis (LDA) with or without the use of Genetic Algorithm (GA) as a variable selection tool with the primary aim to discriminate feeding regimes based on the grass vs those based on the legume, access time (2, 4 or 6 h/d), grazing day (first vs last grazing day during the seven day grazing period) and the milking time (morning vs afternoon morning). The best fitting discriminant models based on FT-MIR spectra were able to correctly predict 100% of the samples differing for the pasture forage, 91.9% of the samples differing for grazing day, and 97.1% of the samples regarding their milking time. The access time to pasture was correctly predicted by the model in 60.3% of the samples, and the classification ability was improved to 77.0% when considering only the 2 h/d and 6 h/d classes.