Prediction accuracies of cheese-making traits using Fourier-transform infrared spectra in goat milk

The objectives of this study were to explore the use of Fourier-transform infrared (FITR) spectroscopy on 458 goat milk samples for predicting cheese-making traits, and to test the effect of the farm variability on their prediction accuracy. Calibration equations were developed using a Bayesian approach with three different scenarios: i) a random cross-validation (CV) [80% calibration (CAL); 20% validation (VAL) set], ii) a stratified CV [(SCV), 13 farms used as CAL, and the remaining one as VAL set], and iii) a SCV where 20% of the goats randomly selected from the VAL farm were included in the CAL set (SCV80). The best prediction performance was obtained for cheese yield solids, justifying for its practical application at population level. Overall results were similar to or outperformed those reported for bovine milk. Our results suggest considering specific procedures for calibration development to propose reliable tools applicable along the dairy goat chain.

Automated summary

This study used FITR spectroscopy to predict cheese-making traits in 458 goat milk samples and tested the effect of farm variability on prediction accuracy. The best prediction performance was achieved for cheese yield solids, indicating practical application potential. The results suggest the need for specific calibration development procedures for reliable tools in the dairy goat chain.