An intelligent molecularly imprinted sensing platform augmented by interval partial least squares for specific...

Mycophenolic acid (MPA), a prevalent mycotoxin, presents significant analytical challenges in complex silage matrices due to limited selectivity and pronounced electrochemical baseline drift. To address this, we developed an intelligent molecularly imprinted sensor (Fe₃O₄-MGO/MIP/GCE) for the sensitive and selective detection of MPA. This sensing platform incorporates Fe₃O₄ modified magnetic graphene oxide (Fe₃O₄-MGO) as a highly conductive nanocomposite substrate, with pyrrole-3-carboxylic acid (Py3C) serving as a dual-functional monomer. Py3C was electropolymerized in the presence of MPA to form recognition sites that also function as signal transducers. Furthermore, to overcome signal instability and subjective interpretation, we integrated a machine learning-assisted approach using a small-window moving average algorithm for automated baseline correction, coupled with interval partial least squares (iPLS) for precise peak identification. This data-driven strategy eliminates manual intervention, enhances signal fidelity, and improves analytical robustness in complex environments. The sensor demonstrated a wide linear range from 7.5 nmol L^-1 to 5 μmol L^-1, with a low detection limit of 2.1 nmol L^-1 (S/N = 3). Recovery rates in real silage samples ranged from 91.5 % to 102.4 %, confirming high accuracy and strong anti-interference capability. This work successfully merges biomimetic molecular recognition with intelligent signal processing, offering a reliable and objective platform for the on-site monitoring of mycotoxins in agricultural systems.