Integrating Machine Learning and Genomic Applications for Antimicrobial Resistance Prediction in Shigella spp., Ontario, Canada

Antimicrobial resistance (AMR) ranks among the top 10 threats to global public health, contributing to approximately 700,000 deaths annually. Shigella spp. infections are typically treated with antibiotics, however, studies indicate that approximately 46% of shigellosis cases exhibit resistance to commonly used antibiotics. The current antimicrobial susceptibility testing (AST) methods have several limitations, including prolonged turnaround times, which can delay appropriate antimicrobial therapy. The latest advancements in rapid and cost-effective whole genome sequencing (WGS) technologies have transformed diagnostic microbiology and microbial surveillance. The primary objective of this Health System Impact Fellowship will be developing AI-driven tools to predict AMR of Shigella spp. using WGS data. More specifically, machine learning algorithms will be used for (a) examining the evolutionary histories of Shigella spp. serotypes, (b) predicting AMR phenotypes and identifying potential genomic regions associated with resistance, and (c) detecting novel genetic markers. These tools not only assist in the immediate clinical care by guiding the selection of the most suitable antibiotic regimen to treat an infection, predicted based on patient clinical and microbiological data, but also contribute to the long-term goal of controlling the spread of infections and AMR.