Building of prediction models by using Mid-Infrared spectroscopy and fatty acid profile to discriminate the geographical origin of sheep milk

- Predictive models of the geographical origin of sheep milk were built.
- Predictive models were based either on fatty acid composition or on MIR spectra.
- Genetic Algorithms were used to select the informative fatty acids and MIR regions.
- All models correctly predict the geographical origin of sheep milk.
- MIR models can routinely be used to attribute the geographical origin of sheep milk.

Geographical authentication of sheep milk is an issue related to the production of cheeses labelled with a Protected Designation of Origin (PDO). To this purpose we investigated both the capability of the fatty acid composition and the capability of the Mid-InfraRed (MIR) spectra of 250 samples of sheep milk (gathered in different areas of the region Sardinia) to discriminate the samples as what regards their geographical origin. Genetic Algorithms (GA) were applied to the fatty acid profile and to the spectra to select the informative variables for developing discriminant models able to correctly classify the samples. The models were validated on unknown samples obtaining correct predictions of 96% using the selected fatty acids and of 99% using the selected MIR spectral regions. For routine control analysis, MIR spectroscopy is preferred for being a non-destructive, cheap and real-time analytical method.