Mass balance of milk components and yield prediction in Caciocavallo Palermitano pasta filata cheese

This study evaluated, through closed mass balances, the partitioning of milk components during the cheesemaking process for the production of traditional pasta filata cheese. In addition, a statistical model was developed to predict cheese yield 24 h after stretching as a function of milk characteristics. The 1-year study was conducted in 3 cheesemaking plants using the traditional system for the production of Caciocavallo Palermitano cheese, carrying out measurements and sampling during bimonthly cheesemaking trials. The proximate composition of milk and of the fractions obtained after rennet coagulation was determined to assess the distribution of water, proteins, fat, ash, and lactose. Mass balance calculations showed closed mass balances greater than 98.7% for all the components analyzed. The average recovery in the pasta filata cheese was 72.0% for proteins and 76.9% for fat, whereas lactose remained predominantly in the whey (86.6%). Integrated ricotta production significantly contributed to the overall recovery of nutrients at the system level. Concurrently, in the 3 cheesemaking plants, bulk milk samples were collected and cheese yields were recorded 24 h after stretching on a weekly basis to build a database for developing a statistical model to estimate yield. Yield at 24 h after stretching was modeled using multiple linear regression. The final model, based on milk protein, fat, and urea contents, explained 82.2% of the observed variability (R2 = 0.822; root mean square error = 0.1769). Leave-one-dairy-out cross-validation confirmed the stability and generalization ability of the model (r = 0.898). The mean absolute error was 0.147, and the mean bias error was negligible. Bland–Altman analysis showed limits of agreement of ~3% of the mean yield.