A machine learning-assisted Cu-MOF/OPD/RB triple-emission ratiometric fluorescence sensing platform for the detection and discrimination of glutathione.

Herein, a machine learning-assisted Cu-MOF/OPD/RB triple-emission ratiometric fluorescence sensing platform has been developed for the highly sensitive and selective detection of glutathione (GSH). The bifunctional Cu-MOF serves as the core, exhibiting intrinsic blue fluorescence at 450 nm and peroxidase-like activity to catalyze the oxidation of o-phenylenediamine (OPD) into fluorescent DAP (545 nm) in the presence of H2O2. Rhodamine B (RB) is further introduced to provide red emission at 580 nm. DAP quenches the fluorescence of the Cu-MOF through an inner filter effect (IFE) while simultaneously enhancing the emission of RB via fluorescence resonance energy transfer (FRET), thereby establishing an interconnected three-channel ratiometric sensing system (F450, F545, F580). GSH inhibits the peroxidase activity of the Cu-MOF via strong chelation of Cu2+ and also directly reduces DAP through its reducing properties, collectively suppressing the generation of DAP. Consequently, F545 and F580 decrease, while F450 recovers due to the weakened IFE. A ratiometric detection method was established based on the signal F450/(F545 + F580), achieving a detection limit of 0.27 μM for GSH. The method exhibited satisfactory recoveries of 90.0%-105.0% with RSD ≤ 5.8% in real samples. Moreover, machine learning models (PCA, K-means) use the three fluorescence intensities as multi-dimensional inputs, allowing accurate GSH quantification and effective discrimination from interfering reducing agents, significantly enhancing selectivity in complex matrices.