FTIR结合化学计量学方法对皱环盖菇总糖和蛋白质含量快速预测

Fast prediction of total sugar and protein content in Stropharia rugosoannulata by means of FTIR and chemometrics

皱环盖菇是近年来深受消费者喜爱的食药用菌新品种,总糖、蛋白质作为其主要的营养成分,含量的高低与其营养价值关系密切。但是,现有含量检测方法步骤复杂,检测时间长,开发简单、快速含量预测方法具有重要意义。本研究利用傅里叶变换红外光谱采集7种不同基质栽培皱环盖菇子实体样品的红外光谱共计420个,对比原始光谱与预处理后光谱,确定光谱预处理方法,采用竞争性自适应重加权算法(CARS)对皱环盖菇子实体总糖、蛋白质所对应的特征波谱进行选择,利用随机森林(RF)、偏最小二乘回归(PLS)、支持向量机(SVM)进行建模,比较建模结果确定总糖含量的最佳预测模型为PLS,其中Rc为0.9928(误差为0.0072),RMSEC为0.9308,Rp为0.9814(误差为0.0186),RMSEP为1.1662,RPD为7.3411;蛋白质含量的最佳预测模型为RF,其中Rc为0.9947(误差为0.0053),RMSEC为0.3803,Rp为0.986(误差为0.014),RMSEP为0.7491,RPD为8.4375。以上结果表明,红外光谱技术结合化学计量法可快速、准确地预测皱环盖菇子实体总糖、蛋白质含量,研究结果可为快速预测总糖、蛋白质含量提供技术支持。

Stropharia rugosoannulata is a new cultivated edible and medicinal mushroom favourably received by consumers in recent years.Its total sugar and protein content,is closely related to its nutritional values.However,the existent content detection methods are complex and time-consuming,and it is greatly significant to develop simple and rapid prediction methods.In this study,a total of 420 infrared spectra from fruiting bodies of S.rugosoannulata cultivated on seven different substrates was collected by Fourier transform infrared spectroscopy FTIR.The spectral pretreatment method was established by comparing the original spectrum and the pretreated spectrum,while the total sugar and protein characteristic spectra were selected by competitive adaptative reweighted sampling(CARS).Random forest(RF),partial least square regression(PLS),and support vector machine(SVM)were used for modeling.The results proved that the best prediction model for total sugar content was PLS,in which Rc was 0.9928(the error was 0.0072),RMSEC was 0.9308,Rp was 0.9814(the error was 0.0186),RMSEP was 1.1662,and RPD was 7.3411;the best prediction model for protein content was RF,with Rc at 0.9947(the error was 0.0053),RMSEC at 0.3803,Rp at 0.986(the error was 0.014),RMSEP at 0.7491,and RPD at 8.4375.The results showed that infrared spectroscopy combined with stoichiometry could quickly and accurately predict the total sugar and protein content of fruiting bodies of S.rugosoannulata.