Machine learning-assisted quadruple-signal Au@Pt nanozyme sensing platform for quantitative detection of ractopamine

Ractopamine (RAC), a β-adrenergic agonist with growth-promoting activity, raises food safety concerns because of potential residues in animal-derived products. An Au@Pt nanozyme-based quadruple-signal lateral flow immunosensor was developed for RAC analysis. Antibody-conjugated Au@Pt nanozymes served as detection probes, while RAC-bovine serum albumin (RAC-BSA) and BSA-modified quantum dots (QDs-BSA) were immobilized on the test line. After chromatographic development, grayscale and fluorescence signals were measured directly from the intact test line. Captured Au@Pt generated the grayscale signal and attenuated QDs-BSA fluorescence through the inner filter effect. The test line was then excised and incubated with 3,3',5,5'-tetramethylbenzidine/hydrogen peroxide to generate catalytic colorimetric and photothermal signals. Signals were acquired by smartphone imaging and handheld infrared thermometry, and multichannel red-green-blue (RGB) data were analyzed using machine-learning (ML) models. The statistically calculated limits of detection were 0.044 ng/mL for the grayscale mode and 0.056 ng/mL for the photothermal mode. For ML-assisted fluorescence and catalytic colorimetric quantification, the ML-derived practical lower limits were 0.008 and 0.02 ng/mL, respectively, corresponding to the lower boundaries of the selected working ranges of 0.008-10 and 0.02-10 ng/mL. Recoveries ranged from 91.75% to 106.96% in pork and from 92.45% to 107.52% in pig urine. Measurements of spiked samples agreed well with enzyme-linked immunosorbent assay results, while authentic RAC-positive samples were comparable to liquid chromatography-tandem mass spectrometry results. This platform provides a multimodal strategy for portable RAC analysis in pork and pig urine.