Feasibility of mid-infrared spectroscopy for the prediction of milk coagulation properties in Alpine and Saanen goats breeds

Milk coagulation process is typically monitored using lactodynamography, which measures rennet coagulation time (RCT, min), curd-firming time as the time to reach a curd firmness of 20 mm (k20, min) and curd firmness after 30 minutes (a30, mm). However, this method has practical limitations, including reliance on skilled personnel, long analysis times, and rennet availability and costs, making large-scale phenotyping challenging. In this context, mid-infrared (MIR) spectroscopy may offer a valuable alternative as a rapid predictive technique. While effective models have been developed for cattle, buffaloes, and sheep, research in goats is limited. This study evaluated the potential use of MIR spectroscopy to predict milk coagulation properties and pH in goat milk. A total of 405 individual milk samples were collected during monthly official milk testing from Saanen and Alpine goats farmed in 8 commercial herds of the Veneto region (Italy). Each sample was divided in two aliquots. The first aliquot was analyzed using CombiFoss™ 7 system for milk quality assessment and MIR spectrum acquisition. The second aliquot was used for the determination of pH (through pH meter) and RCT, k20 and a30 (through lactodynamographic analysis). Milk spectra were matched with their corresponding measured values of pH, RCT, k20 and a30 and a partial least squares analysis was performed. The model was fine-tuned using leave-one-out cross validation; spectral data points were centered and scaled. The number of latent variables was set automatically but limited at a maximum of 15 to avoid overfitting. The data set was randomly split into a calibration set (including 70% of the samples) and a validation set (including 30% of the samples), ensuring similar means and standard deviations across both the sets. The average values recorded were 10.60 min for RCT, 2.87 min for k20, 41.18 mm for a30, and 6.64 for pH. In external validation, the coefficients of determination were 0.62 for k20, 0.67 for RCT, 0.68 for a30 and 0.82 for pH, while the ratio of prediction to deviation ranged from 1.50 (k20) to 2.36 (pH). In conclusion, MIR spectroscopy demonstrated potential as a tool for predicting milk coagulation properties in goat milk and might be implemented in genetics and breeding.