红外光谱结合多元统计分析快速鉴别不同种类牛肝菌

采用傅里叶变换红外光谱结合多元统计分析方法快速鉴别不同种类食用牛肝菌。采集10个不同种类93个牛肝菌子实体的红外光谱,分析食用牛肝菌的红外光谱特征;用多元散射校正(multiplicative signal correction,MSC)、标准正态变量(standard normal variate,SNV)、二阶导数(second derivative,SD)、Norris平滑(ND)、正交信号校正(orthogonal signal correction,OSC)、小波压缩等方法对光谱进行优化处理;经优化处理的光谱数据分别建立马氏距离分类模型及偏最小二乘判别分析(partial least squares discriminant analysis,PLSDA)。结果显示,牛肝菌在3 325、2 934、2 927、1 637、1 547、1 402、1 375、1 259、1 453、1 081、1 029 cm-1等附近有多个吸收峰,主要归属为蛋白质、多糖、氨基酸等的特征吸收峰。MSC+SD+ND(15∶5)和SNV+SD+ND(15∶5)两种预处理方式前10个主成分累积贡献率分别为95.58%、95.54%,基于两种预处理方法建立马氏距离分类模型,验证集预测准确率分别为90%和95%。PLS-DA结果显示经MSC+SD+ND(15∶5)和SNV+SD+ND(15∶5)预处理不易于区分牛肝菌种类;原始光谱经正交信号校正及小波压缩(orthogonal signal correction wavelet compression,OSCW)、优化处理并进行PLS-DA分析,能够很好地区分不同种类牛肝菌。马氏距离分类模型不仅能反映样品的分类情况,同时计算出与测试样品相似度最大的物种,可为食用菌种类鉴别和未知物种鉴定提供可靠依据;OSCW预处理后进行PLS-DA分析能有效鉴别不同种类牛肝菌,为野生食用菌的鉴别分类提供一种辅助方法。

Infrared Spectroscopy Combined with Multivariate Statistical Analysis to Quickly Identify Different Species of Bolete Mushrooms

Fourier transform infrared spectroscopy combined with multivariate statistical analysis was used to establish a
rapid method for the identification of different species of edible bolete mushrooms. The infrared spectral characteristics of 93
bolete samples of 10 different species were analyzed. The original infrared spectra were pretreated by multiplicative signal
correction (MSC), standard normal variate (SNV), second derivative, Norris smooth, orthogonal signal correction (OSC) and
wavelet compression. The optimized spectral data were used to establish a mahalanobis distance classification model and a
partial least squares discriminant analysis (PLS-DA) model. The results showed that the characteristic absorption peaks of
protein, polysaccharide and amino acid appeared at wavenubmers around 3 325, 2 934, 2 927, 1 637, 1 547, 1 402, 1 375,
1 259, 1 453, 1 081, and 1 029 cm?1. The cumulative contribution rates were 95.58% and 95.54% in the PLS-DA model
based on MSC + SD + ND (15:5) and SNV + SD + ND (15:5) pretreatment, respectively. The Mahalanobis distance
classification model was established base on the two pretreatment methods and the prediction accuracies of validation
set were 90% and 95% respectively. Bolete species could not be well distinguished by the PLS-DA model, when the data
were pretreated by the MSC + SD + ND (15:5) and SNV + SD + ND (15:5). PLS-DA analysis of the original spectra after
optimization with orthogonal signal correction wavelet compression (OSCW) could distinguish different species of boletes.
The Mahalanobis distance classification model could reflect the classification of the samples and compute the greatest
similarity with the tested species, which can provide a reliable basis for the classification of edible mushrooms and for the
identification of unknown species. OSCW pretreatment combined with PLS-DA analysis can effectively identify different
species of boletes, providing an auxiliary method for the identification of wild edible mushrooms.