Revealing Antimicrobial Resistance Trends In the Food Chain with Machine Learning Tools
Principal Investigator: Casey Cazer
DESCRIPTION (provided by applicant):
Antimicrobial resistance (AMR) surveillance could reveal relationships between livestock AMR, human AMR infections, and, importantly, stewardship policies. However, incompatible minimum inhibitory concentrations (MICs) arising from varied susceptibility testing protocols, and the lack of bioinformatics tools to investigate multidrug resistance (MDR), limit our ability to answer these fundamental questions. This project’s long-term goal is to identify previously AMR hidden trends across the food chain and characterize the relationship between livestock antimicrobial stewardship and AMR in livestock and humans. This will enable the creation and support of new AMR mitigation policies. These goals align with the priorities of “Mitigating Antimicrobial Resistance Across the Food Chain” (A1366). In this seed grant, we will develop new tools for AMR and MDR trend analysis and pilot them with cattle-associated Escherichia coli data from national AMR surveillance programs. First, we will detect MIC trends and quantify the impact of the 2012 prohibition of extra-label cephalosporin use on cattle-associated E. coli AMR (Objective 1). Random forests, a supervised machine learning approach, will be used to impute missing MICs within and across surveillance programs. MIC distributions will be assessed with Cox proportional hazards models to identify changes over time. Second, we will apply association mining, a novel unsupervised machine learning method, to phenotypic and genomic AMR data to compare E. coli MDR patterns across the food chain (Objective 2). The tools and results generated by this seed grant will support our future proposals to comprehensively investigate the effect of AMR mitigation policies on multiple bacterial species and hosts.