Machine learning-assisted mixed-valence Fe/Cu nanozyme dual-channel sensing platform for high-throughput discrimination of antioxidants

In situ coordination assembly of Cu 2 O/Cu-TCPP(Fe) bimetallic nanozyme. • Intermetallic Cu + /Fe 3+ electron transfer accelerates the peroxidase-like activity. • ML-integrated dual-channel array distinguishes six antioxidants with 100% accuracy. • Precise quantification of antioxidant mixtures in biofluids via RF. Accurate identification and quantitative analysis of antioxidants in complex biological fluids are the keys to the early detection of severe diseases. The single-channel sensors suffer from weak specificity and critical multi-component signal crosstalk, which hardly meet the accurate analytical requirements. Herein, an intelligent colorimetric sensing platform of bimetallic mixed-valence nanozyme was integrated with machine learning (ML) to improve the analytical precision. Fe/Cu nanosheets of mixed valence were successfully acquired through Cu 2 O-coordinated anchoring in situ into iron porphyrin (TCPP(Fe)). This mixed-valence strategy effectively promoted the synergistic catalysis and valence cycling among the metal active sites, thereby significantly enhancing the peroxidase (POD)-like activity of the Cu 2 O/Cu-TCPP(Fe) ( V max as high as 8.107 × 10 −7  M·s −1 and K m as low as 0.913 mM). DFT calculations indicated that the introduction of Cu 2 O promoted strong Cu 3d and Fe 3d orbital hybridization and an upshift of the Cu d-band center, accelerating interfacial electron transfer and Fe/Cu valence cycling, thereby revealing the intrinsic mechanism underlying the enhanced catalytic activity. Taking the differential catalytic responses of the nanozyme with two different substrates, a highly sensitive dual-channel sensor array was designed with machine learning. Combined with algorithms of artificial neural network (ANN), random forest (RF), and support vector machine (SVM), etc., 100% accurate qualitative discrimination of six antioxidants was remarkably completed. Relying on the outstanding capability of the Random Forest (RF) algorithm to remarkable robustness in handling high-dimensional sample datasets, the sensor array successfully established a precise quantitative model for multiple antioxidants in complex samples, remarkably and intelligently sensing the diverse components in urine and serum. This work paved a new way for highly active bimetallic nanozymes to construct a trustworthy sensing platform as a diagnostic tool for the intelligent profiling of multi-component biomarkers in complex matrices. Download: Download high-res image (150KB) Download: Download full-size image