Enhancing surveillance and early warning of infections and antimicrobial resistance using machine learning and deep learning: A systematic review
Jahanzaib Latif, Shijin Zhang, Sadaqat Ur Rehman, Ahsan Wajahat, Ahsan Nazir, Azhar Imran · Neurocomputing · 2026
Antimicrobial resistance (AMR) is a critical global health threat that is projected to cause 10 million deaths annually by 2050. Traditional diagnostic and surveillance methods, which are limited by speed and scalability, are increasingly inadequate against rapidly evolving pathogens. This review evaluates the transformative potential of artificial intelligence (AI), machine learning (ML), and deep learning (DL) in redefining AMR management through enhanced surveillance, drug discovery, and clinical decision-making. A systematic review of 42 peer-reviewed studies (2013–2024) was conducted following PRISMA guidelines. Databases including PubMed, Scopus, and Web of Science were searched using terms like “antimicrobial resistance,” “machine learning,” and “drug discovery.” The inclusion criteria focused on AI/ML/DL applications in AMR, excluding non-English studies and grey literature. Data extraction and bias assessment were performed using standardized tools (QUADAS-2, Cochrane Risk of Bias Tool). AI-driven genomic surveillance tools (e.g., DeepARG and TB-DROP) achieved > 90% accuracy in detecting resistance genes and predicting drug susceptibility, surpassing conventional methods. Temporal surveillance models, such as LSTM networks, achieved 91% accuracy in MRSA outbreak prediction, while genomic prediction tools reduced resistance detection times by 20% compared to conventional methods. Generative adversarial networks (GANs) have accelerated drug discovery, exemplified by halicin, a broad-spectrum antimicrobial effective against multidrug resistant pathogens. Generative AI designs novel antimicrobial candidates, producing 2264 compounds with > 90% in vitro efficacy and compressing early-stage drug discovery timelines. Clinically, reinforcement learning reduced ICU antibiotic misuse by 30% and decision support systems improved pediatric allergy prediction (94% accuracy). AI clinical decision support simulation studies reduced inappropriate antibiotic use by 30% in ICUs and predicted pediatric allergies with 94% accuracy in retrospective validation. Key challenges include dataset biases (overrepresentation of high-income regions), model opacity (only 30% used explainability tools), and translational gaps (15% of AI-discovered compounds reached trials). AI has immense potential for combating AMR through precision diagnostics, proactive surveillance, and accelerated drug development. However, addressing data inequity, computational costs, and ethical concerns is critical. Future efforts must prioritize globally representative datasets, lightweight AI architectures, interdisciplinary collaboration, and ethical frameworks to ensure an equitable implementation. These promising results primarily derive from retrospective studies and computational predictions; prospective clinical validation and real-world implementation evidence remain limited. This review underscores the urgency of responsible innovation in translating AI advancements into real-world lifesaving solutions.