Dual-mode MXene-based biosensor coupled with machine learning: a smart sensing strategy for detection of zearalenone in food.

BACKGROUND: Zearalenone (ZEN) is a potent endocrine disruptor that poses serious health risks through reproductive toxicity. Traditional detection methods, including high-performance liquid chromatography (HPLC), fall short in terms of speed, sensitivity, and portability for on-site food safety monitoring. This underscores the necessity of developing intelligent and field-deployable analytical strategies. RESULTS: We reported a machine learning-enhanced sensing platform based on a dual-modal Ti3C2 MXene biosensor. The nanoprobe orchestrated fluorescence quenching and peroxidase-mimicking colorimetry in a single construct, converting ZEN concentration into two orthogonal optical readouts. The system achieved a detection limit of 0.35 ng mL-1, improving to 0.27 ng mL-1 upon algorithmic fusion. Validation of complex food matrices was conducted using spiked recovery experiments, with recovery rates ranging from 98.90% to 106.91% and a relative standard deviation consistently below 4.84%. Furthermore, in naturally contaminated samples, the results generated by this system were virtually identical to those obtained by high-performance liquid chromatography (HPLC), while significantly reducing analysis time, offering flexible mode selection, and providing cross-validation between the two orthogonal readouts. SIGNIFICANCE AND NOVELTY: By unifying dual-modal transduction with machine-learning-driven data integration, this platform delivered sensitivity and accuracy surpassing conventional approaches while retaining field-deployable simplicity. It provided a robust solution for ZEN detection in real-world food samples, addressing a critical gap in rapid, on-site mycotoxin monitoring and advancing intelligent biosensing paradigms.