Calibrated Machine Learning Using Whole Genome Sequencing

This study was designed as a retrospective; secondary analysis of publicly available bacterial whole-genome sequencing (WGS) data linked to antimicrobial susceptibility phenotypes. The primary objective was to develop and validate machine learning (ML) models that predict antimicrobial resistance (AMR) from WGS-derived features, with an explicit public health framing: enabling near–real-time genomic surveillance of resistance patterns, supporting outbreak investigations via rapid resistance profiling, and informing antimicrobial stewardship through population-level risk stratification