Abstract Surface-enhanced Raman spectroscopy (SERS) was examined to explore the feasibility of SERS technique to develop a rapid, non-destructive, and reliable spectroscopic method for qualitative and quantitative analysis of chlortetracycline (CTC) and oxytetracycline (OTC) in animal feed. Study samples were prepared by spiking tetracycline-free animal feed at different concentration ranges. In several Raman shift regions including characteristic peaks, spectral variation and Raman intensity difference among CTC and OTC groups at different concentrations were clearly visualized, depending on the type of tetracycline. The k-nearest neighbor (KNN) and linear discriminant analysis (LDA) models yielded excellent correct classification rates while showing no or only one misclassification of spiked samples as false-negative. The first two canonical variables in the chemometric modes for classification accounted for more than 95% variation in SERS spectra. Of the models developed for predicting CTC and OTC concentrations, multiple linear regression (MLR) and partial least squares regression (PLSR) models for CTC quantification showed outstanding model performance and ability, with coefficient of determination (r2) (>0.94), low predictive error rate (

中文翻译：

---

使用表面增强拉曼光谱 (SERS) 快速检测和预测动物饲料中的金霉素和土霉素

摘要 表面增强拉曼光谱 (SERS) 被检查以探索 SERS 技术的可行性，以开发一种快速、无损、可靠的光谱方法，用于动物饲料中金霉素 (CTC) 和土霉素 (OTC) 的定性和定量分析。通过在不同浓度范围内添加不含四环素的动物饲料制备研究样品。在包括特征峰在内的几个拉曼位移区域中，不同浓度的 CTC 和 OTC 组之间的光谱变化和拉曼强度差异清晰可见，具体取决于四环素的类型。k-最近邻 (KNN) 和线性判别分析 (LDA) 模型产生了极好的正确分类率，同时没有或只有一个错误分类为假阴性的加标样本。用于分类的化学计量模式中的前两个典型变量占 SERS 光谱中超过 95% 的变化。在为预测 CTC 和 OTC 浓度而开发的模型中，用于 CTC 量化的多元线性回归 (MLR) 和偏最小二乘回归 (PLSR) 模型显示出出色的模型性能和能力，决定系数 (r2) (>0.94)，预测性低错误率（