Behavior-based modeling of Listeria monocytogenes growth in dairy products: A meta-analytical framework for food safety

The microbiological safety of the dairy chain remains a critical public health concern given the resilience of Listeria monocytogenes . This study presents a large-scale meta-analytical study of 867 growth profiles extracted from the ComBase database, covering a wide range of matrices from fluid milk to long-ripened cheeses. After a rigorous filtering process, a final curated dataset of 306 high-quality kinetic profiles was used for behavioral clustering and modeling. Using a Random Forest variable-importance and sensitivity-analysis framework, we ranked the relative contribution of the evaluated predictors. Temperature consistently emerged as the most influential variable across the tested predictor sets, while pH and matrix-composition descriptors contributed to secondary variability. Through Hierarchical Clustering on Principal Components (HCPC), the dairy landscape was segmented into three distinct behavioral risk zones that transcend commercial nomenclature. Correlation heatmaps revealed a “kinetic fingerprint” where thermal dominance in milk (r = 0.90) shifts to a multifactorial control in cheese (r = 0.36). Furthermore, we observed limitations associated with the exclusive use of the Scott reference strain, where environmental isolates frequently grow faster in fluid matrices. Cluster-specific secondary models improved the description of matrix-dependent kinetic behavior, supporting behavior-based segmentation as a complementary framework for dairy risk assessment under the conditions represented in the curated dataset.