Machine learning-assisted robust identification of kanamycin under acidic fluctuations using a self-regulated nanoconfined sensor.

The ultrasensitive detection of kanamycin is important for acidic-condition analytical sensing. However, pH variations in acidic samples often trigger significant background interference, leading to false positive or negative results. In this study, a machine learning-assisted dual-probe nanoconfined sensor (KBA&C4/SiNWs/GMP) was developed for the precise identification of kanamycin under acidic fluctuations. By co-functionalizing a glass micropipette (GMP) filled with silica nanowires (SiNWs) with a kanamycin-binding aptamer (KBA) and a proton-responsive DNA strand (C4 DNA), the sensing interface integrates target recognition with pH-responsive interfacial regulation. To enhance recognition reliability, machine learning algorithms, including Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Linear Discriminant Analysis (LDA), were employed to classify the ionic current responses. The machine learning models accurately classified concentration-dependent ionic current responses despite signal variations caused by acidity fluctuations, enabling the sensor to maintain a low limit of detection (7.31 fM) and a high classification accuracy (94.17%) in acidic matrices. This work provides a promising strategy for reliable kanamycin detection in acidic and complex sample environments.