甘蓝型油菜茎秆纤维组分含量和木质素单体G/S近红外模型构建

【目的】探索近红外光谱分析技术在甘蓝型油菜茎秆纤维组分含量及木质素单体G/S测定中应用的可能性。【方法】采集近红外光谱,根据马氏距离GH(Global H)筛选出103份纤维组分含量材料和75份木质素单体G/S材料作为定标样品,采用Van Soest法和GC-MS法对茎秆纤维组分含量和木质素单体比例进行测定,统计结果表明定标样品化学测定值变异范围较大,3次重复差异较小,可用于近红外模型构建。运用不同光谱预处理方法和化学计量学方法建立校正模型,对比各模型性能参数,筛选出最优定标模型并用检验集对模型进行验证。【结果】采用修正偏最小二乘法(MPLS)建立模型最佳。中性洗涤纤维(neutral detergent fiber,NDF)、酸性洗涤纤维(acid detergent fiber,ADF)、酸性洗涤木质素(acid detergent lignin,ADL)与木质素单体G/S的交叉验证相关系数(1-VR)分别为0.864、0.861、0.872和0.920,定标相关系数(RSQ)分别为0.892、0.891、0.907和0.953。用检验集对模型进行验证,NDF、ADF、ADL及木质素单体G/S模型的外部检验相关系数(RSQ)分别为0.837、0.818、0.870和0.935,其预测标准差(SEP)为0.680、0.636、0.348和0.054。【结论】试验所建模型质量较好,能快速测量茎秆纤维组分含量和木质素单体G/S,可为油菜抗病抗倒伏育种研究提供技术支持。

Construction of a Near Infrared Model for Detecting Stem Fiber Component Content and Lignin Monomer G/S in Rapeseed

【Objective】The aim of this study is to explore the feasibility of rapidly detecting the content of fiber composition and monomer G/S with near-infrared spectroscopy (NIRS) in rapeseed(Brassica napus L.). 【Method】 After the NIRS was collected, 103 samples for fiber component content and 75 for monomer G/S in rapeseed were selected by removing outlier and redundant samples according to measure distance. Fiber component content was measured with Van Soest method and lignin monomer G/S was measured with GC-MS. Subsequently, data analysis showed that the little difference for the chemical test, and the range of variation in the calibration samples was wide enough for building the NIRS models. By using the optimum spectrum pretreatment and regression method along with internal cross validation, the calibration models were established, and the model accuracy was tested. 【Result】The best regression method for prediction of fiber component and lignin monomer G/S in rapeseed was modified partial least-squares (MPLS) by using near infrared spectroscopy. The cross-validation correlation coefficient (1-VR) of NIRS model for neutral detergent fiber (NDF), acid detergent fiber (ADF), acid detergent lignin (ADL) and monomer G/S model was 0.864, 0.861, 0.872 and 0.920, respectively, and the determination coefficient (RSQ) was 0.892, 0.891, 0.907 and 0.953, respectively; The results of testing the validation set showed that external validation correlation coefficient (RSQ) was 0.837、0.818、0.870 and 0.935, respectively, and square errors of prediction (SEP) was 0.680, 0.636, 0.348 and 0.054, respectively. 【Conclusion】These results indicated that the model established in our research had relatively high accuracy in prediction tests for stem fiber component content and lignin monomer G/S in rapeseed, which could provide support for rapeseed disease and lodging resistance in research and breeding project.