Predicting S. aureus antimicrobial resistance with interpretable genomic space maps

Antimicrobial resistance (AMR) is a global healthcare threat, and rapid antibiotic treatment selection is crucial. Generative topographic mapping (GTM) is a non-linear dimensionality reduction method used to visualize sequence space and predict AMR. Applying to AMR data from over 5000 S. aureus isolates, GTM models showed reasonable accuracy for all drugs. GTMs allow antibiotic-wise comparison of resistant phenotypes and analyze genetic determinants responsible for drug resistance.