MACHINE LEARNING AND GENOMICS FOR IMPROVED FOOD SAFETY AND RISK ASSESSMENT OF SALMONELLA ENTERICA IN CHICKEN

Abstract

Salmonella enterica is a leading cause of foodborne illnesses worldwide and is commonly associated with poultry. Salmonella has many closely related serovars, yet these serovars exhibit significant variability in many characteristics including host range, virulence, growth behavior, stress response, and antimicrobial resistance. In the past, this intricate and dynamic population heterogeneity of Salmonella severely hampered control efforts, but, today, this has improved through the sequencing of Salmonella genomes. Whole genome sequencing (WGS) provides a better understanding of the evolutionary and ecological adaptations that underlie the survival of Salmonella against antimicrobials, oxidative agents, non-optimal temperatures, and other stressors in the environment and their hosts. Coupling machine learning with WGS expands on these advantages by enabling the identification of genetic patterns that may not be immediately apparent. The overall goal of this research was to explore how machine learning and genomics can be integrated to improve food safety. First, a machine learning model was developed to identify stress response genes in Salmonella isolated from different poultry processing stages. It was found that beyond genes encoding for cold and heat shock proteins, other genes involved in lipopolysaccharide biosynthesis, DNA repair and replication, and biofilm formation are involved in Salmonella’s overall stress response mechanism. Additionally, a machine learning model was developed to predict antimicrobial resistance (AMR) phenotypes in Salmonella isolates using WGS data. The model predictions were comparable to existing bioinformatic methods for AMR prediction and identified AMR genes that are typically not the resistance determinants public health agencies focus on. Expanding this approach for AMR surveillance could lead to the discovery of novel AMR genes. Lastly, a quantitative microbial risk assessment for Salmonella in chicken that incorporated Salmonella heterogeneity in growth and virulence was developed. The findings revealed that variations in virulence have a greater impact on the risk of salmonellosis than variations in growth rate. Overall, this research contributes to efforts to enhance food safety measures and reduce chicken-associated Salmonella illnesses.