近红外光谱法在蛹虫草诱变筛选及其发酵条件优化中应用

应用近红外(NIR)光谱技术结合化学计量学法建立测定蛹虫草菌丝体中腺苷、蛋白、多糖和虫草酸4种主要有效成分含量的定量分析模型,可应用于蛹虫草高产突变株的筛选和发酵条件的优化。468株蛹虫草突变株在不同的摇瓶发酵条件下进行发酵,收集菌丝体粉末样品并采集NIR光谱,同时采用常规方法分别测定样品的4种有效成分含量。偏最小二乘法(PLS)用于建立NIR光谱与有效成分含量间相关模型。采用蒙特卡罗偏最小二乘法(MCPLS)识别异常样本和选择合适的校正集样本数;采用移动窗口偏最小二乘法(MWPLS)筛选波长变量;以逼近度(Da)为考察指标,选择最有效的光谱预处理方法和最适的PLS隐变量数,最终得到测定蛹虫草菌丝体中腺苷、蛋白、多糖和虫草酸含量的最优PLS模型,其校正集样本参考值与预测值间相关系数(Rc)分别为0.92943,0.98479,0.90785和0.85131;预测集误差(RMSEP)为0.66714,0.02065,0.01131和0.01159,表明模型具有很好的拟合度和预测性能,应用该方法进行蛹虫草诱变筛选和发酵条件的优化是可行的。

Application of near infrared spectroscopy in screening Cordyceps militaris mutation strains and optimizing their fermentation process

To fast screen high-yield Cordyceps militaris mutations strains and optimize their fermentation process, near infrared (NIR) spectroscopy technology combined with chemometrics has been applied to establishing models for simultaneous determination of adenosine, protein, polysaccharide and Cordyceps militaris acid contents in Cordyceps militaris powder samples. Fermentations were implemented in Erlenmeyer flask with 468 Cordyceps militaris mutations strains under various fermentation conditions and Cordyceps militaris powder samples were collected. Then their NIR spectra were obtained using UV-VIS-NIR spectrometer and their adenosine, protein, polysaccharide and Cordyceps militaris acid contents were determined using reference methods. Partial least squares (PLS) method was employed to model the relationships between NIR spectra and the above mentioned components' contents in Cordyceps militaris powder samples. Monte Carlo partial least square (MCPLS) was applied to identify the outliers and select suitable number of calibration samples. Moving window partial least square (MWPLS) was applied to select the characteristic wavelength of the components. The degree of the approaching (Da) was employed as criterion for selecting effective pretreatment methods investigated. The optimum models for determination of adenosine, protein, polysaccharide and Cordyceps militaris acid contents in Cordyceps militaris powder samples were obtained with the above mentioned optimization. Their correlation between actual and predictive values of calibration samples (Re) was 0.92943, 0.98479, 0.90785, and 0.85131, respectively. Their root mean square error of prediction set (RMSEP) was 0.66714, 0.02065, 0.01131, and 0.01159, respectively. The obtained results demonstrated that the fitting and the predictive accuracy were satisfactory. It is feasible to apply this method to screen the high yield Cordyceps militaris mutation strains and optimize their fermentation process.