amR: an R package suite to predict antimicrobial resistance in bacterial pathogens

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Antimicrobial resistance (AMR) is a complex trait with critical implications for diagnostics and treatment.
• Current computational methods for AMR identification often lack interpretability and fail to explore cross-species or multi-drug patterns.
• Understanding AMR mechanisms across diverse bacterial species and drug classes is essential for effective intervention.

Purpose of the Study:

• To develop an integrated R package suite, amR, for comprehensive and interpretable bacterial AMR prediction.
• To provide a framework for identifying AMR mechanisms from bacterial genome data across multiple molecular scales.
• To enable exploration of cross-species and multi-drug AMR patterns.

Main Methods:

• The amR suite comprises three modular packages: amRdata for data curation and feature extraction, amRml for training interpretable machine learning models, and amRviz for interactive visualization.
• Genomic and antimicrobial susceptibility testing data are processed to extract multi-scale features (gene clusters, protein domains, COGs, ResFinder features, structural variants).
• Interpretable machine learning models are trained per species-drug combination, with feature importance analysis for mechanism discovery.

Main Results:

• The amR suite successfully predicts AMR, achieving a median Matthews Correlation Coefficient of 0.89 across 23 drugs for Shigella sonnei.
• The framework enables identification of AMR mechanisms across species and drugs by analyzing multi-scale genomic features.
• The amRviz package facilitates interactive exploration of model performance, feature importance, and cross-model patterns.

Conclusions:

• The amR R package suite offers an accessible and comprehensive framework for bacterial AMR research.
• Interpretable machine learning models within amR facilitate hypothesis generation and discovery of AMR mechanisms.
• amR supports improved diagnostics and treatment strategies by providing insights into bacterial resistance patterns.