Machine Learning Approaches for Predicting Antimicrobial Resistance from Genomic Data: A Systematic Review

##article.abstract##

Antimicrobial resistance (AMR) represents a major global health challenge, significantly
limiting the effective prevention and treatment of bacterial infections. Advances in whole-
genome sequencing (WGS) have enabled rapid identification of genetic determinants
associated with resistance. However, conventional rule-based approaches primarily rely on
previously characterized resistance genes and mutations, restricting their ability to detect
novel or complex resistance mechanisms. In this context, machine learning (ML) has
emerged as a powerful alternative, leveraging high-dimensional genomic datasets to
model complex and nonlinear relationships between genotype and phenotype. Over the
past decade, ML-based approaches have been increasingly applied to predict antibiotic
susceptibility directly from genomic information in clinically important pathogens such as
Mycobacterium tuberculosis, Escherichia coli, Staphylococcus aureus, Klebsiella
pneumoniae, and Salmonella species. This systematic review summarizes recent
advancements in genomic feature extraction, algorithm development, model validation
strategies, and the clinical applicability of ML-driven AMR prediction tools. Various
computational frameworks, including supervised, unsupervised, and deep learning models,
are critically examined alongside benchmarking practices, validation methodologies, and
model interpretability techniques. Emerging directions such as explainable artificial
intelligence, federated learning, and multi-omics integration are also discussed for their
potential to enhance predictive performance and clinical utility. Despite promising results
in controlled settings, several challenges remain, including data heterogeneity, limited
generalizability, regulatory constraints, and ethical considerations. Overall, this review
provides a comprehensive perspective on the current state and future prospects of ML-
based genomic diagnostics for antimicrobial resistance.