已核实 DOI 相关产品: 2

Detection of veterinary drug residues in Tan sheep based on fluorescence hyperspectral response analysis: A new strategy combining 2D correlation spectroscopy and density functional theory

Journal 2026 · IF 7 · 发表于 2026 年 7 月 20 日
作者: Fujia Dong,Hui Li,Yuxia Hu,Rongguang Zhu,Zongxiu Bai,Ying Xu,Xingxing Wang,Weiguo Yi,Songlei Wang
阅读全文 在 PubMed 查看

摘要

The presence of veterinary drug residues seriously endangers human health. Rapid and accurate detection is a necessary measure to reduce the pollution caused by veterinary drug residues. An interpretable strategy combining two-dimensional correlated spectroscopy (2D-COS) with density functional theory (DFT) was proposed for the analysis of fluorescence hyperspectral imaging (FHSI) responses of quinolone residues and in situ detection. The experiment took danofloxacin (DFX) as an example, the fluorescence hyperspectral data of 200 samples were collected using a 280 nm excitation wavelength. Studies have shown that the main emission peak of the DFX experimental spectrum is located at 441 nm. To uncover the intrinsic features under the overlapping signals, the 2D-COS was utilized to effectively resolve the hidden potential characteristic peak at 415.50 nm. Furthermore, it was revealed that the first excited state (S1) of the DFX molecule originated from the electron transition from the HOMO to the LUMO using TD-DFT, and the theoretical vertical transition wavelength was 416.29 nm. The theoretical predicted values were in excellent agreement with the 2D-COS analytical values, which proved the physical authenticity of the spectral fingerprints. Multivariate data processing methods were utilized to construct linear, nonlinear and ensemble learning regression models. The results showed that the 2D-COS-SVR model preprocessed with the second-order derivative had the best performance (with R2 P being 0.9388, RMSEP being 0.0783, and RPD being 4.042). The SHAP method was introduced to conduct an interpretability analysis on the optimal model, which confirmed that 415.50 nm was the key feature with the highest contribution to the prediction model, which was consistent with the results of DFT calculations. An integrated framework of "theoretical calculation-spectral analysis-model interpretation" was established in this study, which provided a mechanism-driven and interpretable new paradigm for the detection of antibiotic residues in meat.

本研究中使用的阿拉丁产品

2 款产品在本文中被引用