Machine learning-assisted colorimetric sensor array using a microbially synthesized Prussian blue nanozyme with
Fang Chen1, Mengping Li1, Xinqi Xu1
1Laboratory of Micro & Nano Biosensing Technology in Food Safety, Hunan Provincial Key Laboratory of Food Science and Biotechnology, College of Food Science and Technology, Hunan Agricultural University, Changsha, 410128, China.
None:
Sulfur-containing metal salts (SCMs) are widely used as food additives, but their excess residues or chemical interconversion during processing may generate more toxic species, posing significant health risks. However, conventional analytical methods for SCMs suffer from high cost, time-consuming procedures, and reliance on sophisticated instrumentation. Herein, we report a nanozyme-based colorimetric sensor array constructed from a microbial synthesized Prussian blue nanozyme (Bio-PB) with intrinsically enhanced peroxidase-like activity. The array integrates Bio-PB with three chromogenic substrates (TMB, OPD, ABTS) to generate a three-channel cross-reactive sensing platform. Upon exposure to SCMs, competitive consumption of reactive oxygen species (ROS) by SCMs with different sulfur oxidation states produces unique colorimetric fingerprints. Critically, machine learning (linear discriminant analysis, hierarchical cluster analysis, and principal component analysis) is employed to decode the multi-channel response patterns, transforming subtle cross-reactive differences into statistically robust classification. This machine learning-assisted approach achieves 100% discrimination accuracy of five SCMs (Na2S, Na2S2O3, Na2S2O4, Na2S2O5, Na2S2O8) and correctly identifies 98.78% of 82 unknown samples. It further enables semi-quantification over 10-180 μM, distinguishes binary to quinary SCM mixtures, and maintains excellent performance in complex food matrices (milk, red wine, and eggs). Notably, unlike conventional nanozyme relying solely on redox interaction, Bio-PB hybrid leverages specific binding of SCMs to bacterial membrane components via electrostatic/hydrophobic interactions, endowing high selectivity and robust stability in real-world matrices. With its simplicity, rapidity and cost-effectiveness, coupled with machine learning-powered pattern recognition, this bio-PB based sensor array offers a promising platform for on-site SCMs monitoring and food safety screening.
