Predicting antimicrobial resistance for precision medicine

Clinical Snapshot

Key Findings

• P Value: Not reported

• Effect Size: Not applicable — no original quantitative data reported; effect sizes from cited primary studies are not pooled or formally summarised

• Primary Outcome: Conceptual synthesis demonstrating that ML/AI-based AMR prediction using WGS data, informed by mechanistic understanding of resistance evolution, can theoretically enable a precision medicine approach to bacterial infection — targeting pathogens specifically while limiting microbiome disruption and resistance evolution

• Nnt Or Sensitivity: Not reported in this perspective; individual ML model performance metrics (sensitivity, specificity, accuracy for AMR prediction) are discussed qualitatively from cited primary studies but not systematically synthesised or tabulated

• Confidence Interval: Not reported — no meta-analytic or pooled estimates provided

Clinical Application

Currently limited by WGS turnaround times in acute settings (typically 24–72 hours for clinical WGS pipelines, compared to hours for rapid phenotypic susceptibility tests), computational infrastructure requirements, need for curated reference databases, and the gap between genotypic prediction and phenotypic resistance expression. Feasibility is improving with nanopore sequencing and cloud-based bioinformatics but remains unproven at scale in routine clinical practice. Australia faces a significant and growing AMR burden, with the ACSQHC and the Australian Government's National Antimicrobial Resistance Strategy 2020–2025 prioritising surveillance and stewardship. The Australian Genomics initiative and state-based public health laboratory networks (e.g., NSW Health Pathology, Victorian Infectious Diseases Reference Laboratory) are expanding WGS capacity. However, WGS-based AMR prediction is not yet embedded in routine clinical workflows or PBS-funded diagnostic pathways. The TGA regulates in vitro diagnostic devices, and any ML-based AMR prediction tool would require TGA conformity assessment before clinical deployment. RACGP and IDSA-equivalent guidance (Australasian Society for Infectious Diseases, ASID) does not yet incorporate ML-based AMR prediction into empirical therapy algorithms. Narrow-spectrum novel therapeutics discussed in the review (e.g., phage therapy, bacteriocins, targeted antivirulence agents) are largely investigational in Australia, with phage therapy available only on a compassionate access basis via the TGA's Special Access Scheme. Patients with confirmed or suspected bacterial infections in settings where WGS-based diagnostics are available or being implemented; particularly relevant for infections caused by priority AMR pathogens (e.g., ESKAPE organisms, Mycobacterium tuberculosis, carbapenem-resistant Enterobacterales)

Abstract

Antibiotics are among medicine's greatest successes, but resistance evolution threatens their continued efficacy. Decades of research have deepened our understanding of the mechanisms and evolutionary dynamics of antimicrobial resistance. More recently, advances in machine learning (ML) and artificial intelligence (AI) show promise in predicting antimicrobial resistance in pathogens based on rapid whole-genome sequencing and other accessible data. In this perspective, we highlight advances in understanding the mechanisms and spread of antimicrobial resistance. We discuss how this knowledge, coupled with ML- and AI-based approaches, can inform the prediction of resistance and a precision-medicine strategy that targets pathogenic bacteria specifically, thereby limiting resistance evolution and collateral damage to the microbiome. These accurate predictions of bacterial vulnerabilities will enable the adaptation of classical antimicrobial treatments with adjuvants, as well as the use of novel, narrow-spectrum therapeutics. Implementing these strategies, while also identifying key challenges, will help bring this strategy into clinical practice.

References

1. 1.Fink, T., Rybniker, J., & Bollenbach, T. (2026). Predicting antimicrobial resistance for precision medicine. Cell Host & Microbe. https://doi.org/10.1016/j.chom.2026.05.031