Antibiotic Impacts and Resistance Gene Dissemination in Anammox-Based Wastewater Treatment: Insights from Machine-Learning-Assisted Data Synthesis

This review introduces a machine-learning-assisted framework to systematically evaluate antibiotic impacts on anammox-based treatment systems, with emphasis on low-strength nitrogenous wastewater, based on the currently available literature. A meta-analysis integrated with unsupervised machine learning, including Uniform Manifold Approximation and Projection (UMAP) and Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN), was applied to reveal hidden patterns in reactor performance, antibiotic concentrations, and antibiotic resistance gene (ARG) dynamics. Macrolides and tetracyclines exerted the strongest inhibition, reducing nitrogen removal efficiency (NRE) up to approximately 34% at concentrations of 1–100,000 μg/L, while sulfonamides and fluoroquinolones caused milder impacts. UMAP-HDBSCAN analysis revealed distinct clustering patterns by antibiotic type, class, ARGs type and resistance mechanism, and highlighted a data-driven transition in system performance across substrate concentration ranges, with intermediate conditions (around 260 mgN/L) marking the shift between lower and higher NRE levels. Stress from both tetracycline-class antibiotics and nontetracycline antibiotics promoted the proliferation of >80% of tetracycline-related ARGs (e.g., tetA, tetG, tetM, tetX), while macrolide- (ermF) and sulfonamide-related (sul1) ARGs showed partial reductions. Resistance mechanisms associated with efflux pumps and antibiotic deactivation were strongly linked to ARG dissemination. By embedding machine learning into an environmental review context, this study demonstrates how data-driven synthesis can advance mechanistic understanding and support risk-informed management of anammox-based treatment systems.