Machine learning-integrated portable sensing platform based on COF-OMe@Valine-MnO2 nanozyme for sensitive detection of organophosphorus pesticides.

The extensive use of organophosphorus pesticides (OPs) in agriculture has raised critical concerns regarding environmental safety and public health. However, accurate and interference-resilient detection of OPs in complex matrices remains a formidable analytical challenge. Herein, we employed hydrothermal synthesis coupled with an in situ doping strategy to rationally engineer a porous COF-OMe@Valine-MnO2 nanozyme exhibiting intrinsic phosphatase-mimicking activity. The nanozyme enabled direct electrochemical detection of two representative OPs, namely chlorpyrifos and phoxim, without enzymatic amplification, by relying on its oxidative hydrolysis of the organophosphorus substrates. The results demonstrated wide linear ranges of 1-150 and 0.05-10 μM, low limits of detection of 14.2 and 3.35 nM, and sensitivities of 2.23 and 9.48 μA/(μM·cm2) for chlorpyrifos and phoxim, respectively. To overcome conventional signal ambiguity, we incorporated machine learning (ML) to simulate the electrochemical reaction dynamics and optimize detection parameters, thereby achieving high classification accuracy across concentration levels and enabling the predictive recovery of analyte signals under varying matrix conditions. The resulting platform was successfully applied to raw rice grains and agricultural irrigation water samples, yielding spike recoveries of 89.6%-106.2%, which strongly correlated with results obtained from high-performance liquid chromatography with ultraviolet detection. This study not only establishes an ML-accelerated nanozyme sensor for reliable OP screening in complex real-world samples but also offers a generalizable framework for intelligent, on-site monitoring of hazardous substances.