Enhanced Food-borne Pathogen Surveillance through Whole- Genome Sequencing and Machine Learning for Novel Threat Detection in One Health Framework

Foodborne pathogens present a significant global public health challenge, with millions of cases reported annually, resulting in severe illnesses and fatalities. The growing complexity of food production and distribution, alongside the rising threat of antimicrobial resistance (AMR), has intensified the need for advanced surveillance systems. The One Health framework, which underscores the interdependence of human, animal, and environmental health, demands integrated and innovative approaches to address these challenges. Whole-Genome Sequencing (WGS) has emerged as a highly effective tool in foodborne pathogen surveillance, offering unparalleled resolution and comprehensive genetic insights that surpass traditional methods. Despite the widespread adoption of WGS technology by many countries and regions, the effectiveness of current WGS surveillance methods is limited by existing reference databases and the boundaries of human knowledge. There is a pressing need to improve the speed, accuracy, and reliability of WGS in responding to unrecognized or underappreciated threats. This thesis aims to develop new bioinformatics workflows and analytical methods that reduce dependency on existing databases and knowledge bases, enhancing the ability to identify and assess novel threats within the One Health framework. Chapter 1 provides a comprehensive overview of the evolution of pathogen detection methods, tracing the development from traditional laboratory techniques to molecular methods and finally to bioinformatics approaches. The chapter focuses on summarizing the current WGS analysis techniques and the application of machine learning (ML). Chapter 2 constructs a WGS workflow for the surveillance of foodborne pathogens, which includes core genome SNP phylogenetic clustering, analysis of antimicrobial resistance genes, and virulence gene profiling. The chapter focuses on a comprehensive whole-genome analysis of 90 foodborne Salmonella isolates collected from Colombia, leading to the identification of two novel multi-drug resistance plasmids. Chapter 3 applies the WGS workflow developed in Chapter 2 to conduct an in-depth genomic study of three multi-drug resistant Listeria innocua Previously considered non-pathogenic, L. innocua had not been a primary target of surveillance efforts. This chapter presents a detailed comparison of the resistance gene islands and virulence factors found in these isolates, specifically examining the relationship between their resistance genes and the LIPI-4 virulence gene cluster found in pathogenic Listeria monocytogenes of serotype 4b. The study offers insights into the horizontal gene transfer of risk genes between Listeria species, emphasizing the potential for cross-species transmission of genetic elements that could elevate food safety risks. Chapter 4 explores the application of machine learning (ML) in predicting biocide and metal tolerance in two non-typhoidal Salmonella (NTS) serotypes. This chapter develops a stacked classifier model based on WGS data and applies it to the genotypic analysis and phenotypic prediction of biocide and metal tolerance in Salmonella Indiana and Salmonella Thompson. hapter 5 investigates the study of taxonomical identification and genomic characterization of two dulcitol-positive novel subspecies of Cronobacter dublinensis isolated from infant cereals in China. The chapter details the isolation process, phenotypic characterization, and whole-genome sequencing of these subspecies, offering insights into their unique genetic features. The study also examines the metabolic pathways related to dulcitol utilization, contributing to a better understanding of the ecological niches occupied by these pathogens.