Meat is nutritious and highly susceptible to contamination by foodborne pathogens, spoilage bacteria and other harmful microorganisms, so that its food safety is of great concern. In recent years, machine learning methods have been extensively employed in the field of food safety. In this paper, we review the application of machine learning methods from two perspectives: detection and predictive modeling of harmful microorganisms in meat. This review also analyzes the limitations of machine learning methods at the present stage, and provides an outlook on their prospects in meat microbial safety.

Foodborne pathogenic bacterial biofilm is a spatially structured multi-cellular community that is adhered to food contact surfaces. Conventional disinfection methods cannot effectively prevent and control bacterial biofilm due to its strong adhesiveness and stress resistance, causing food safety hazards and a serious effect on consumer health. In this article, recent studies on the formation mechanism and control strategies of foodborne pathogenic bacterial biofilm are reviewed. The roles of the functions of bacterial long-chain appendages, quorum sensing system and extracellular polymeric substances (EPSs) in the formation of bacterial biofilm are summarized and discussed based on improved biofilm adhesiveness, changed bacterial state, and enhanced stress resistance. This review divides the existing strategies for biofilm prevention and control into three categories: physical, chemical, and biological, and analyzes the principle, advantages and disadvantages of each category in order to provide a theoretical reference for the development of more efficient strategies for biofilm prevention and control in the food field to achieve better prevention and control of food microbial safety.

Microbial communities interact in a complicated way, which can have a detrimental effect on the host’s health either individually or collaboratively. By employing causal inference methods to analyze the characteristics of the relationship between the microbiota and host observation data, the mechanism of action of microorganisms on host health can be inferred, which will help to reduce the possibility of causing unfavorable health consequences. This article reviews recent progress in the application of causal inference in the field of microbiological research, introduces the conception and development of causality, describes the paths of causal inference, modeling and interpretation, and summarizes the causal inference methods and their applications in various fields especially in the microbiological field. However, the mechanism of the causality between microbial communities and their external environments is not yet understood fully. Applying causal inference methods in research on microbial communities will be a hot topic in the future. Thus, the theory and methods of causal inference need be further refined to elucidate the interaction mechanism of microbial communities and their causality with human health.

Contamination of spoilage and pathogenic microorganisms in food substrates can negatively influence food quality and threaten food safety. The application of physicochemical antimicrobial strategies can effectively inhibit the growth and reproduction of harmful microorganisms, but may also cause irreversible negative effects on food quality. Some antagonistic yeasts alone or combined with antimicrobial agents can effectively control harmful microorganisms. This review systematically and comprehensively summarizes the existing literature on the antimicrobial effects and mechanisms of yeasts and their derivatives from the above two aspects. The sources, species and mechanism of action of antagonistic yeasts as well as their antimicrobial effects when combined with other antimicrobial technologies are discussed. Additionally, the advantages and limitations of yeast as a natural drug carrier combined with various chemical antimicrobial agents are analyzed. This review aims to provide a scientific reference for developing new yeast-based antimicrobial agents, which will be helpful in improving food quality and ensuring food microbial safety.

The viable but nonculturable (VBNC) state refers to a state in which bacteria lose their ability to grow and reproduce in conventional media under adverse environmental stress conditions. Bacteria in the VBNC state are usually still metabolically active and remain pathogenic and virulent. Bacteria in the VBNC state can recover and return to a state where they can be cultured normally under certain suitable conditions, posing potential risks to food safety. In this review, the morphological, physiological, biochemical and biomacromolecular characteristics of bacteria, especially foodborne pathogens, in the VBNC state are comprehensively described. Meanwhile, various factors that induce bacteria to enter the VBNC state, including temperature and pH unsuitable for bacteria survival, are summarized. The laboratory conditions for bacterial recovery from the VBNC state are outlined with special reference to adequate nutrition as an indispensable factor for recovery from the VBNC state. Finally, the major methods for the detection of bacteria in the VBNC state are discussed, including molecular biological methods, flow cytometry, and optical microscopy. Molecular biological methods have been widely used in the detection of bacteria in the VBNC state due to their advantages of accuracy, rapidity and high sensitivity. This review provides more ideas for further research on the induction, resuscitation and detection of foodborne pathogens in the VBNC state.

The ability of Listeria monocytogenes to survive under nutrient-deprived conditions may significantly impact L. monocytogenes viability in food and food safety. This study aimed to investigate the viability of L. monocytogenes against detergents in different nutrient states and the relevant survival mechanism. Key observations revealed an increased tolerance of L. monocytogenes to Sodium Dodecyl Sulfate (SDS) following nutrient stress, notably in the 1/100 TSB-YE (Tryptic Soy Broth-Yeast Extract) group. The injury rate of the 1/100 TSB-YE group was significantly lower than that of other nutrient groups (P < 0.05). Besides, morphological adaptations to nutrient stress were characterized by increased cell length and volume, accompanied by an upsurge in the secretion of extracellular substances. Transcriptome analysis revealed significant gene expression changes in response to nutrient dilution and starvation, with a notable number of differentially expressed genes (DEGs) involved in membrane components, ATP binding, and metabolic processes, which were validated through the RT-qPCR tests. This study demonstrates that nutrient stress can impede the growth of L. monocytogenes, but also induces a cross-protection response against SDS. These findings could be supplemented with its inactivation parameters to the risk assessment system and the response to stress was provided a theoretical reference for the prevention and control of L. monocytogenes.

The growth and survival characteristic of Salmonella Enteritidis under acidic and osmotic conditions were studied. Meanwhile, a probabilistic model based on the theory of cell division and mortality was established to predict the growth or inactivation of S. Enteritidis. The experimental results demonstrated that the growth curves of planktonic and detached cells showed a significant difference (p<0.05) under four conditions, including pH5.0+0.0%NaCl, pH7.0+4.0%NaCl, pH6.0+4.0%NaCl, and pH5.0+4.0%NaCl. And the established primary and secondary models could describe the growth of S. enteritis well by estimating four mathematics evaluation indexes, including determination coefficient (R²), root mean square error (RMSE), accuracy factor (A_f) and bias factor (B_f). Moreover, sequential treatment of 15% NaCl stress followed by pH 4.5 stress was the best condition to inactivate S. Enteritidis in 10 h at 25 °C. The probabilistic model with Logistical or Weibullian form could also predict the inactivation of S. Enteritidis well, thus realize the unification of predictive model to some extent or generalization of inactivation model. Furthermore, the primary 4-parameter probabilistic model or generalized inactivation model had slightly higher applicability and reliability to describe the growth or inactivation of S. Enteritidis than Baranyi model or exponential inactivation model within the experimental range in this study.