Machine Learning-Assisted Nanozyme–Porphyrin Dual-Channel Sensor Array for Glycosaminoglycan Pattern Recognition

Abstract Selective discrimination of glycosaminoglycans (GAGs) is crucial for drug safety and quality control but remains challenging because GAGs share similar disaccharide repeating units and overlapping polyanionic features. Nanozyme-based sensor arrays offer a cross-reactive fingerprinting strategy, yet many require multiple nanozymes with single readouts and correlated responses, increasing operational complexity and limiting pattern separability. Herein, we constructed a dual-channel sensor array by assembling two cationic near-infrared (NIR) porphyrins with a multifunctional Pt–Ni/rGO nanozyme that integrated efficient fluorescence quenching and oxidase-like catalysis. Upon GAG binding, each sensing element generated dual outputs, including NIR fluorescence response driven by competitive association between anionic GAGs and cationic porphyrins, and characteristic UV–vis absorption changes arising from Pt–Ni/rGO-catalyzed TMB oxidation to TMBox followed by TMBox–GAG assembly. Machine learning analysis of the four cross-reactive signals generated by the two sensing elements enabled 100% accurate discrimination of hyaluronic acid, heparin, dextran sulfate, and chondroitin sulfate in PBS over 25–500 μg/mL. Its practical utility was further demonstrated by identifying trace GAG contaminants in Hep down to 1%, classifying unknown samples, and discriminating GAGs in serum. This work expands the analytical utility of multifunctional nanozymes and provides a simple strategy for fingerprint-based GAG analysis with enhanced signal dimensionality and separability.