Multifrequency Electrochemical, Aptamer-Based Sensor Fingerprints Enable Machine Learning Identification and Quantification of Cross-Reactive Targets

Cross-reactivity remains a persistent challenge in small-molecule biosensing, including the use of recognition elements such as aptamers, which often interact with structurally similar small-molecule targets. In this work, we overcome the limitation of poor specificity of small-molecule-binding aptamers to build electrochemical aptamer-based (E-AB) sensors through coupling the frequency-dependent signature sensor response (current) obtained from continuous square-wave voltammetry (cSWV) with machine learning analysis to distinguish signals between aptamer interactions with structurally similar targets. Employing sensors fabricated with an aminoglycoside aptamer, known to exhibit a signal in the presence of aminoglycosides, including tobramycin and kanamycin, we observe appreciable and similar sensor responses (recorded as signal gain) in the presence of either molecule using traditional voltammetric methods. However, because the binding to these targets leads to distinct aptamer-target structures, we demonstrate here that the time-resolved, multifrequency electrochemical interrogation capabilities of cSWV provide distinct frequency-dependent sensor responses for each target. These distinctive signal patterns build the basis of our approach, in which we couple the signal pattern for each cross-reactive target with a machine learning model (random forest), enabling target classification and subsequent quantification across multiple electrodes, yielding an overall classification accuracy of 0.80 and robust regression performance (R2 = 0.91 for tobramycin and R2 = 0.93 for kanamycin). To our knowledge, this is the first demonstration of combined E-AB sensors with cSWV machine learning in a unified framework to qualitatively and quantitatively detect cross-reactive targets, establishing a powerful framework that enhances biosensing reliability and effectively addresses distinguishing cross-reactive targets.