Predictive modeling of the potentially hazardous psychrotolerant Bacillus cereus group in fried rice

Bacillus cereus group (BCG) includes closely related bacterial species with different phenotypic characteristics, such as pathogenic potential, enzymatic capacity, and thermotypes. Psychrotolerant BCG (pBCG) strains can grow and produce toxins at low temperatures, creating concerns for the food industry. However, many current routine food diagnosis methods do not consider pBCG, and predictive modeling, which is an essential tool for food safety and public health, has not been developed using pBCG, but rather using mesophilic strains (mBCG). Given the limited information on predictive modeling and accurate strain identification of pBCG, we developed a predictive model for fried rice, a known causative food for BCG foodborne disease, using pBCG food isolates and whole-genome sequencing for accurate taxonomic classification. In predictive modeling, four pBCG food isolates selected through phenotypic screening grew at temperatures above 5 degrees C, whereas mBCG reference strains did not grow below 13 degrees C. The primary and secondary models for pBCG (covering 5-37 degrees C) and mBCG (covering 13-37 degrees C) fit well, with R-2 > 0.98. By validating the dynamic model under three time-varying temperature profiles, we observed root mean square error values < 0.42 log CFU/g and acceptable simulation zone values > 82%. The four pBCG isolates used in the predictive model were identified using whole-genome sequencing as B. cereus sensu stricto, B. toyonensis, and B. mycoides, which carried enterotoxin genes. Psychrotolerant signatures of the 16S rRNA and cspA were detected in the BCG9 isolate. The predictive model and genomic characterization of pBCG strains can be applied to manage and control pBCG, ensuring the food quality and safety of fried rice products in the cold chains.