采用近红外漫反射光谱进行牛樟芝菌丝体组分检测的研究

目的：采用偏最小二乘法结合近红外光谱技术建立分析牛樟芝发酵菌丝体中多糖和三萜含量的泛化能力强且预测精度高的定量分析模型，以满足牛樟芝原料药及其相关产品实际检测中的应用。方法：通过扫描化学诱变和液体深层发酵获得的165个牛樟芝菌丝体样品获得近红外光谱，采用常规方法测定样品中多糖和三萜的含量。应用蒙特卡罗偏最小二乘法识别异常样品并确定校正集样品数量为基础，以逼近度（Da）为评价指标，采用可移动窗口偏最小二乘法对特征波长变量、最佳光谱预处理方法及建立模型的重要参数进行筛选。结果：最终确定牛樟芝菌丝体中多糖和三萜的校正集和预测集样品实验测定值与预测值间相关系数（R_c和R_p）分别为0.931 1 g·g^-1和0.995 9 g·g^-1、0.927 9 g·g^-1和0.943 4 g·g^-1；均方根误差（RMSEC和RMSEP）分别为0.031 33 g·g^-1和0.010 72 g·g^-1、0.031 34 g·g^-1和0.012 02 g·g^-1。结论：该模型具有很好的预测性能。

Objective:Near infrared spectroscopy combined with partial least squares was used to establish quantitative model for polysaccharide and triterpene in mycelia of Antrodia camphorata with strong generalization ability and high prediction precision,which could satisfy the practical test of raw materials and related products of Antrodia camphorata.Methods:NIR spectra of 165 samples treated by chemical mutagenesis and liquid fermentation were obtained and the contents of polysaccharide and triterpene were determined by conventional methods.Monte Carlo partial least squares was applied to identify abnormal samples and to define the number of calibration samples.Then the characteristic wavelengths,the spectral pretreatment methods and the key modeling parameters were optimized based on the degree of approximation(Da)by moving window partial least square filter.Results:The calibration and prediction correlation coefficient(R_c and R_p)for polysaccharide were 0.931 1 g·g^-1 and 0.995 9 g·g^-1,respectively.And those for triterpene were 0.927 9 g·g^-1 and 0.943 4 g·g^-1,respectively.The root mean square errors of calibration and prediction(RMSEC and RMSEP)were 0.031 33 g·g^-1 and 0.010 72 g·g^-1 for polysaccharide and 0.031 34 g·g^-1 and 0.012 02 g·g^-1 for triterpene,respectively.Conclusion:The models had good prediction performance.