火炬松针叶儿茶素含量近红外预测模型的建立  

Establishment of near-infrared prediction model for catechin content in Pinus taeda L.needles

作　　者：吕欣欣 冯志恒 林熙 李赛楠 黄少伟 LYU Xinxin;FENG Zhiheng;LIN Xi;LI Sainan;HUANG Shaowei(Guangdong Key Laboratory for Innovative Development and Utilization of Forest Plant Germplasm,College of Forestry and Landscape Architecture,South China Agricultural University,Guangzhou,Guangdong 510642,China)

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资源创新与利用重点实验室,广东广州510642

出　　处：《西北农林科技大学学报（自然科学版）》2022年第6期28-34,共7页Journal of Northwest A&F University(Natural Science Edition)

基　　金：“十三五”国家重点研发计划项目(2017YFD0600502-3)。

摘　　要：【目的】建立火炬松针叶儿茶素含量近红外预测模型,为选育高生物活性成分火炬松良种奠定基础。【方法】以102个火炬松单株的针叶为试验材料,利用液相色谱-质谱联用技术(LC-MS)测定其儿茶素含量。使用近红外成分分析仪采集样品的近红外光谱信息,对采集的光谱信息采用不同的方式(一阶导数(FD)、标准正态变量转换法(SNV)、平滑算法、乘积分散校正法(MSC)和标准化预处理以及FD+SNV、MSC+FD)进行预处理,结合偏最小二乘法建立回归模型,比较不同预处理方法建立的回归模型参数,选择最佳光谱预处理方法,建立火炬松针叶儿茶素含量近红外预测模型,并对模型的预测准确性进行验证。【结果】FD+SNV为最佳的近红外光谱信息预处理方法;建立了火炬松针叶儿茶素含量的近红外预测模型,该模型的主成分数为14,校正集相关系数(R_(C))为0.9696,校正集均方根误差(RMSE_(C))为1.3084,交互验证集相关系数(R_(V))为0.8171,交互验证集均方根误差(RMSE_(V))为3.1052。经过外部验证,验证集火炬松针叶样品的儿茶素含量实测值与预测值有显著相关性(R=0.8807)。【结论】建立了火炬松针叶儿茶素含量近红外预测模型,该模型可以准确、高效地预测火炬松针叶的儿茶素含量。【Objective】The near infrared prediction model of catechin content in needle leaves of loblolly pine was established to lay a foundation for breeding improved loblolly pine varieties with high bioactive components.【Method】The needles of 102 loblolly pine individual plants were selected for determination of catechin content by liquid chromatography-mass spectrometry.The near-infrared spectral information was collected by near-infrared component analyzer.The collected spectral information was preprocessed in different ways(first derivative(FD),standard normal variable transformation(SNV),smoothing algorithm,multiplication integral dispersion correction(MSC),standardized preprocessing,FD+SNV and MSC+FD).Combined with partial least square method,the regression model was established,the parameters established by different pretreatment methods were compared and the best spectral pretreatment method was selected.Then,the near-infrared prediction model of catechin content in needle leaves of loblolly pine was established and the prediction accuracy of the model was verified.【Result】FD+SNV was the best preprocessing method of infrared spectrum information.The near-infrared prediction model of catechin content in needle leaves of loblolly pine was established with principal component score of 14.The correlation coefficient of the correction set(R_(C))was 0.9696,the root mean square error of the correction set(RMSE_(C))was 1.3084,the correlation coefficient of the interactive verification set(R_(V))was 0.8171,and the root mean square error of the interactive verification set(RMSE_(V))was 3.1052.After external verification,the actual value of catechin content in needle leaf samples of Pinus taeda had significant correlation with the predicted value(R=0.8807).【Conclusion】The established near-infrared prediction model can accurately and efficiently predict catechin content in needles of loblolly pine.

关 键 词：火炬松 儿茶素 近红外预测模型 松针 

分 类 号：S791.255[农业科学—林木遗传育种]