Machine learning-enabled microfluidic ratiometric fluorescence sensor array based on lanthanide-gold nanoclusters for visual detection of multicomponent antibiotics.

The accumulation of multiple antibiotics in the environment poses severe threats to ecosystems and human health, demanding rapid, sensitive, and portable detection methods. This study introduces a lanthanide-gold nanocluster (AuNCs)-based microfluidic ratiometric fluorescence sensor array integrated with machine learning for visual multicomponent antibiotic monitoring. Three probes (Eu3+-Lys-AuNCs, La3+-Arg/ATT-AuNCs, Tb3+-BSA-AuNCs) targeting oxytetracycline (OTC), levofloxacin (LEVO), and norfloxacin (NFX) leverage "antenna effects" and structural interactions to generate distinct RGB fingerprint patterns in a capillary-driven microfluidic chip, enabling precise discrimination. The system achieved ultra-low detection limits of 12.48 nM (OTC), 5.63 nM (LEVO), and 10.95 nM (NFX), with linear ranges of 0.1-10 μM, 0.1-2 μM, and 0.1-8 μM, respectively. Furthermore, machine learning algorithms (linear discriminant analysis, confusion matrix or decision tree) enable excellent classification accuracy for single antibiotics, binary mixtures, and ternary mixtures in environmental samples. Therefore, this user-friendly platform, by integrating microfluidics, ratiometric fluorescence, and portable technologies with visual data and intelligent algorithms, provides a scalable solution to the challenge of simultaneous multi-component antibiotic detection in complex environmental matrices.