Artificial Intelligence for Predicting Antimicrobial Resistance and Optimizing Therapy in Escherichia coli Infections

Antimicrobial resistance (AMR) in Escherichia coli is a growing global threat, yet conventional susceptibility testing delays targeted therapy, forcing reliance on empirical broad-spectrum antibiotics. Artificial intelligence (AI) and machine learning (ML) offer transformative solutions by integrating whole‑genome sequencing (WGS) data, clinical parameters, and local resistance patterns to predict resistance phenotypes and optimize antibiotic selection. Recent studies demonstrate that deep learning architectures, including discriminative position‑fused classifiers and attention‑based models, can accurately predict resistance to multiple antibiotics from genomic features, often exceeding ۹۰% accuracy. Beyond prediction, ML frameworks now support empirical and organism‑targeted therapy recommendations, while reinforcement learning explores treatment policies that minimize resistance emergence. Despite these advances, challenges in data generalizability, model interpretability, and prospective validation remain. This review synthesizes current evidence, compares methodological approaches, and outlines a roadmap for clinical integration. AI‑driven AMR prediction promises to shift practice from empirical prescribing toward rapid, individualized, evidence‑based therapy, potentially curbing unnecessary broad‑spectrum use and slowing resistance evolution.

کلیدواژه ها:

Antimicrobial Resistance ، Escherichia coli ، Machine Learning ، Whole‑Genome Sequencing ، Antibiotic Stewardship ، Precision Medicine