Train a drug resistance prediction model

Antimicrobial resistance (AMR) is increasing the risk of treatment failure and prolonging illness because clinicians often must choose antibiotics before susceptibility results are available. This project will use existing resistance phenotype data (MICs and S/I/R interpretations) linked to bacterial genomes to train and validate models that predict resistance from genomic features. By standardizing phenotype records across testing platforms and standards, applying stringent quality control, and benchmarking performance across species, drugs, and settings, we aim to deliver reliable early predictions that support faster effective therapy and better patient outcomes. The models will also enable more precise prescribing and timely de-escalation, strengthening antimicrobial stewardship by reducing unnecessary broad-spectrum use. Aggregated predictions can enhance surveillance by identifying emerging resistance patterns and high-risk lineages, informing outbreak detection and public health response. Finally, a reproducible workflow can complement routine laboratory testing, improve efficiency, and extend decision support to resource-limited settings, strengthening health systems and supporting more equitable AMR management.