Deep Learning Sensing Platform for Simultaneous Detection of Kanamycin and Kanamycin B Based on Dual-Recognition and Catalytic Modulation Mechanism of Aptamer-Functionalized MIL-101 Nanozyme

Abstract

Extensive antibiotic use has raised concerns over residues in food, necessitating rapid and accurate detection. This study integrates smartphone-assisted deep learning with aptamer-functionalized MIL-101(Fe) nanozymes to build a dual-recognition biosensor capable of simultaneously identifying and quantifying kanamycin (KAN) and kanamycin B (KMB). MIL-101(Fe) exhibited strong adsorption energies and high Langmuir-fitted capacities (28.39 mg/g for KAN; 31.36 mg/g for KMB), indicating monolayer binding at Fe─OH sites that partially suppressed catalytic activity. Aptamer modification enabled selective molecular recognition and accelerated electron transfer, synergistically enhancing nanozyme performance. Coupled with a ResNet-based multitask model, smartphone imaging achieved accurate classification of single and mixed samples and precise quantification over 5–1000 nM. Tests in honey and beef showed recoveries of 84.71%–111.50%, confirming stability and practicality. Overall, the aptamer-enhanced MIL-101(Fe) nanozyme combined with deep learning provides a sensitive, portable approach for on-site monitoring of antibiotic residues in food.