目的 比较滇重楼Paris polyphylla var.yunnanensis不同生长年限不同产地红外光谱特征，建立产地鉴别模型。方法 收集13个产地3～8年生栽培滇重楼样品，共计841株个体；通过衰减全反射-傅里叶变换红外光谱仪（attenuated total reflection-Fourier transform infrared spectra，ATR-FTIR）采集样品红外光谱；结合多元散射校正（multiplicative scattering correction，MSC）、一阶导数（first derivative，FD）、二阶导数（second derivative，SD）和三阶导数（third derivative，TD）筛选最佳光谱预处理方式；利用偏最小二乘判别模型（partial least squares discrimination analysis，PLS-DA）鉴别相同生长年限不同产地的样品。结果 三阶导数与多元散射校正结合是对3年生和5～8年生样品进行产地鉴别的最佳光谱预处理方式；二阶导数与多元散射校正结合是鉴别4年生不同产地样品最佳光谱预处理方法。变量投影重要性（Variable importance for the projection，VIP）分析显示，3700～1800 cm^-1在不同PLS-DA判别模型中均为贡献较大的变量波段；当剔除无关信息波段之后，PLS-DA模型能够准确鉴别相同年限不同产地滇重楼根茎样品，训练集和预测集鉴别准确率均为100%；置换检验结果显示模型鲁棒性较好。结论 研究建立的方法能够快速有效鉴别不同产地栽培滇重楼，为不同栽培年限滇重楼产地溯源提供理论基础与方法依据。

Objective To establish geographical discrimination model by comparing spectral characters of Paris polyphylla var. yunnanensis (PPY) with different cultivated locations and different growth period. Methods A total of 841 samples of PPY growing 3—8 years were collected from 13 geographical origins. The sample powder was scanned by attenuated total reflection-Fourier transform infrared spectrometer (ATR-FTIR) for obtaining spectral information. Subsequently, the optimum methods of spectral pretreatment were selected by using multiplicative scattering correction (MSC) combined with first derivative (FD), second derivative (SD) and third derivative (TD). The partial least squares discrimination analysis (PLS-DA) was established using these spectra after selecting different spectral pretreatment methods for geographical discrimination under the same growth period. Results The results indicated that TD coupled with MSC was the best pretreatment approach for origin identification of 3-year old and 5—8-year old samples, while SD coupled with MSC was the best preprocessing approach to identify 4-year-old samples from different habitats. Variable importance for the projection (VIP) analysis showed that the ATR-FTIR spectral band distributions in the whole range of 3700—1800 cm^-1 were important variables for building discrimination models. Moreover, PLS-DA was an effective way for the discrimination analysis of samples from different geographical origins with different cultivation years, showing 100% accuracy rates of training set and prediction set. Furthermore, the permutation test indicated that these models were robust.Conclusion In general, the discrimination method built in the study could effectively discriminate cultivated PPY samples from different geographical origins, which can provide a theoretical foundation and methodology for geographical traceability of different growth periods of PPY.

R286

国家自然科学基金项目（81460584）；云南省重大科技专项（202002AA100007）