Safety-by-Design of Functional Camel Milk Kefir Using Predictive Microbiology and Machine Learning for Salmonella Control

Introduction: Camel milk is a traditional dairy matrix of growing global interest, and kefir fermentation represents a promising strategy for valorizing this resource as a functional dairy product. However, the microbiological safety of camel milk kefir under variable fermentation and storage conditions remains insufficiently characterized. This study aimed to develop an integrated predictive framework for Salmonella inactivation in camel milk kefir by combining classical predictive microbiology, response surface methodology, and machine learning.

Methods: Camel milk was collected in the Souss-Massa region of Morocco, pasteurized at 63 °C for 30 min, and fermented using locally sourced kefir grains. A Doehlert uniform shell design involving three factors and 16 experimental runs was applied to evaluate the effects of kefir grain concentration (1–10% w/v), fermentation time (18–30 h), and storage temperature (4–25 °C) on Salmonella survival. Samples inoculated at approximately 5–6 log CFU/mL were monitored during storage. Physicochemical parameters, including pH, titratable acidity, and Salmonella counts were measured throughout the experiment. Bigelow and Weibull models were used to describe inactivation kinetics, while Random Forest, XGBoost, and Support Vector Regression models were developed for data-driven prediction.

Results: The Weibull model outperformed the Bigelow model in 15 of the 16 runs, with R² values reaching 0.995 and lower AIC and RMSE values. RSM models adequately described the effects of processing factors on Weibull parameters, with higher factors levels associated with greater Salmonella inactivation. Among machine learning models, Random Forest showed the best predictive performance.

Conclusions: This integrated mechanistic and data-driven approach provides a promising safety-by-design framework for optimizing camel milk kefir fermentation and storage conditions to improve pathogen control.