Machine learning-assisted fluorescent sensor array based on siloxane-derived porous polymers for high-throughput discrimination of multiple antibiotics in milk.

Both simultaneous monitoring of multiple antibiotic residues for public health security and accurate single-analyte quantification for regulatory compliance are critically important, though the former remains challenging due to structural diversity among analytes. Herein, we constructed a four-channel fluorescent sensor array based on siloxane-derived porous polymers for discriminative detection of 12 antibiotics spanning four categories (chloramphenicols, nitrofurans, nitroimidazoles, and tetracyclines), and further achieved discrimination of 6 antibiotics in real milk samples at lower concentrations (0.5 μM), along with quantitative determination of oxytetracycline (OTC) in two brands of milk. The array primarily generates differential fluorescence responses through synergistic inner filter effects and photoinduced electron transfer, featuring rapid response (<15 s), high sensitivity (detection limit down to 0.16 μM), and excellent recyclability (POSS-Ph exhibits > 90% fluorescence retention for tetracycline (TC) after 5 consecutive cycles). Integrated with machine learning algorithms (e.g., LDA and KNN), the sensor achieved 100% identification accuracy in spiked milk samples, demonstrating its potential as a robust platform for high-throughput antibiotic screening in complex matrices.