基于近红外光谱技术建立火炬松针叶表儿茶素含量的预测模型  

Prediction Model of L-Epicatechin Content in Pinus taeda Based on Near Infrared Spectroscopy

作　　者：李赛楠 吕欣欣 林熙 冯志恒 毛积鹏 黄少伟 刘天颐 Li Sainan;Lu Xinxin;Lin Xi;Feng Zhiheng;Mao Jipeng;Huang Shaowei;Liu Tianyi(Key Laboratory of Innovation and Utilization of Forest Plant Germplasm Resources in Guangdong Province,College of Forestry and Landscape Architecture,South China Agricultural University,Guangzhou,510642)

机构地区：[1]华南农业大学林学与风景园林学院,广东省森林植物种质创新与利用重点实验室,广州510642

出　　处：《分子植物育种》2023年第11期3761-3770,共10页Molecular Plant Breeding

基　　金：国家重点研发项目子课题(2017YFD0600502-3)资助。

摘　　要：表儿茶素是松针提取物中黄酮类成分的一种,具有较强的抗氧化和清除自由基能力,为了建立一种快速、高效测定火炬松(Pinus taeda L.)针叶表儿茶素(L-Epicatechin)含量的近红外光谱预测模型,使用近红外光谱分析仪在950~1 650 nm的光谱范围内,采集了106个火炬松松针样本的光谱数据,结合液相色谱质谱联用仪对松针中表儿茶素含量的测定结果,采用偏最小二乘法(PLS),建立火炬松针叶表儿茶素的近红外光谱预测模型。结果表明,多元散射校正(MSC)+一阶导数(FD)+滤波拟合(S-G)为最佳光谱预处理方法,当主成分数为15时,表儿茶素含量预测模型可达到最优,其校正相关系数RC=0.973 9,校正均方根误差(RMSEC)=1.198 1,交互验证集相关系数RV=0.804 5,交互验证集均方根误差RMSEV=3.184 8。外部验证集测定值和模型预测值之间具有极显著的相关性(R=0.813 6),预测均方根误差RMSEP=3.629 9。模型RC和RV均最大,RMSEC和RMSEV均较低,说明所建立模型的预测性能较好。预测值与测量值具有高度正相关,模型的预测准确度较高,可以满足对火炬松针叶表儿茶素含量进行快速预测的要求。L-Epicatechin is one of the flavonoids in pine needle extract,which has antioxidant and free radical scavenging abilities.In order to establish a rapid and efficient method for the determination of Pinus taeda L.Near-infrared spectroscopy prediction model of L-Epicatechin content in needles.The spectral data of 106 Pinus taeda needles were collected by Near-infrared spectroscopy in the spectral range of 950~1650 nm.Combined with the results of liquid chromatography mass spectrometry,the regression model was established by partial least squares(PLS)to establish the Near-infrared spectral prediction model of L-Epicatechin in Pinus taeda needles.The results showed that the Multiplicative Scattering Correction(MSC)+First Deviation(FD)+Savitzky golay(S-G)was the best spectral preprocessing method,and the L-Epicatechin content prediction model could reach the optimal when the principal score was 15,with a correction set correlation coefficient RC=0.9739,a corrected root mean square error(RMSEC)-1.1981,a cross-validation set correlation coefficient R=0.8045,and a cross-validation root mean square error RMSEV-3.1848.There was a highly significant correlation between the measured and model predicted values for the external validation set(R=0.8136),with a root mean square error of prediction RMSEP=3.6299.The RC and RV were the largest,and RMSEC and RMSEV were low,indicating that the prediction performance of the developed model was good.The predicted value was highly positively correlated with the measured value,and the prediction accuracy of the model was high,which can meet the requirements for rapid prediction of L-Epicatechin content in Pinus taeda needles.

关 键 词：火炬松 表儿茶素 近红外光谱技术 偏最小二乘法 

分 类 号：S791.255[农业科学—林木遗传育种]