An AI-assisted smartphone platform for H2O2 and glutathione detection using a dual-functional Pd-ReS2-MXene nanoprobes

Abstract

Glutathione (GSH) and H<inf>2</inf>O<inf>2</inf> are important biomarkers in oxidative stress management, and their imbalance is linked to cancer, neurological illnesses, and metabolic diseases. However, interference from structurally identical antioxidants makes it difficult to detect them selectively and sensitively. To overcome this problem, we developed a defect-engineered Pd-ReS<inf>2</inf>-MXene nanozyme with improved peroxidase (POD)-mimetic activity for rapid and accurate biomolecular sensing. Atomically dispersed Pd nanoparticles on ReS<inf>2</inf>-MXene nanosheets, produced through controlled chemical reduction, prove higher catalytic efficiency than standard POD-like catalysts. Compared to Pd-ReS<inf>2</inf> (674.8 U mg−1) and Pd-MXene (897.2 U mg−1), the nanozyme exhibits POD-like activity of 1377.6 U mg−1. For 3,3′,5,5′-tetramethylbenzidine (TMB), Michaelis-Menten kinetics show increased catalytic activity with a V<inf>max</inf> of 11.22 μM s−1 and a K<inf>m</inf> of 0.112 mM. Colorimetric experiments show that Pd-ReS<inf>2</inf>-MXene catalyzes the H<inf>2</inf>O<inf>2</inf>-mediated oxidation of TMB, resulting in a blue product. When GSH reverses the process, TMB reverts to its colorless state. The nanozyme has linear detection ranges of 1–500 nM for H<inf>2</inf>O<inf>2</inf> and 0.05–110 μM for GSH, with detection limits of 0.43 and 1.2 nM, respectively. A paper-based analytical device integrated with a smartphone analyzer allows in situ H<inf>2</inf>O<inf>2</inf> quantification in live cells. GSH detection in serum, plasma, and murine liver tissue lysates confirmed recoveries of 98.4–104 % with low relative standard deviations (1.4–3.4 %) and minimal interference from other amino acids. Pd-ReS<inf>2</inf>-MXene presents a scalable nanozyme platform for biomedical applications such as disease monitoring and therapeutic interventions, with outstanding reusability, stability, and potential for machine learning-assisted real-time sensing. © 2025 Elsevier B.V.