Orthogonal Photoelectrochemical–Colorimetric Fingerprints for Machine Learning-Assisted Discrimination of Four Organophosphorus Pesticides

Abstract Organophosphorus pesticides (OPs) are widely used in agriculture but pose severe neurotoxic threats to human health. Conventional OPs detection assays rely on the inhibition of acetylcholinesterase (AChE) activity by suppressing thiocholine (TCh) generation from acetylthiocholine hydrolysis. However, structurally similar OPs often exhibit overlapping inhibition responses, making the AChE inhibition assay insufficient to distinguish them. Herein, based on the bifunctional CuBi2O4 aerogel nanozyme, a machine learning-assisted orthogonal discrimination strategy is proposed by converting the AChE inhibition response of OPs into photoelectrochemical–colorimetric (PEC-CL) fingerprints for identification of four OPs. The PEC-CL dual-signal outputs are realized through catalytic oxidation and photoelectric conversion processes, while the orthogonal sensing relies on the dual regulatory functions of TCh. In the PEC channel, TCh acts as an interfacial electron donor to accelerate charge transfer, thereby regulating the photocurrent generation. In the CL channel, TCh competes with the chromogenic substrate for reactive oxygen species, thereby modulating the chromogenic reaction. This orthogonal regulation enables a single enzymatic inhibition event to simultaneously produce two independent analytical responses, thereby generating characteristic fingerprints for different OPs. Through linear discriminant analysis (LDA), the coupled PEC-CL responses are transformed into discriminative feature maps, enabling accurate classification and concentration prediction of structurally similar OPs over a linear range of 0.01–5 μg/mL with low detection limits of 1.57–3.46 ng/mL. The sensor achieves 100% discrimination accuracy of four OPs and enables reliable identification in real agricultural samples. By integrating orthogonal dual-mode sensing with supervised machine learning, this strategy offers a promising avenue for intelligent and practical pesticide residue screening and risk assessment in complex samples.