平菇是我国大宗食用菌品种之一, 产量位居我国食用菌的第三位。 平菇不仅味道鲜美, 而且含有丰富优质的蛋白质及具有多种生物活性的平菇多糖, 深受消费者喜爱。 市场上的平菇产品众多, 质量参差不齐, 并且现有营养成分分析方法耗时长、 成本高, 难以满足平菇等大宗食用菌的营养成分检测需求。 傅里叶变换红外光谱(FTIR)技术具有检测速度快、 操作方便、 可同时分析多种化合物、 安全环保等特点, 将其与化学计量学结合, 构建数学模型, 能满足对平菇等大宗农产品营养成分的快速检测、 分析及评价。 以平菇为研究对象, 在全国范围内收集主栽平菇样品85份, 分别进行红外光谱扫描, 并运用多元散射校正(MSC)、 标准正态变换(SNV)、 正交信号校正(OSC)、 光滑加一阶导数(F-G D)、 光滑加二阶导数(S-G D)等5种光谱数据预处理方法, 通过比较模型验证集回归系数, 确定平菇蛋白质模型最佳预处理方式为OSC结合S-G D, 平菇多糖模型最佳预处理方式为OSC结合F-G D。 在最佳光谱预处理下, 采用LASSO算法对7458个光谱波段进行特征波段提取, 获得平菇蛋白质特征波数93个, 平菇多糖特征波数92个, 压缩率达到98%。 将特征波数与化学方法检测得到的平菇蛋白质、 多糖含量值进行拟合, 建立PLS模型。 结果显示, 蛋白质模型校正集回归系数R2为0.999 8, RMSECV为0.047 7, 验证集回归系数R2为0.987 2, RMSEP为0.506 8, RPD为8.840 6大于3； 多糖模型校正集回归系数R2为0.999 9, RMSECV为0.020 1, 验证集回归系数R2为0.980 3, RMSEP为0.292 9, RPD为7.119 8大于3, 模型拟合效果均较好, 预测能力及稳健性良好。 该研究为傅里叶变换红外光谱技术在食用菌营养成分含量快速预测方法的建立提供参考, 为平菇产品的营养品质评价的建立提供基础, 促进平菇乃至其他食用菌产品的优质化发展。

Pleurotus ostreatus is one of the wide varieties of edible fungus, ranking third for its yield in China. Except for its delicious taste, appreciated by consumers, it is known to be rich in high-quality protein and polysaccharides with various biological activities. However, there are different kinds of P. ostreatus following their quality, and the existing nutrient composition analysis methods are time-consuming and high in composition. It is difficult to meet the requirements of the detection of their nutrient composition, as well as for other edible fungi. Fourier Translation Infrared Spectroscopy (FTIR) technology, characterized by high-speed detection, convenient technique, simultaneous analysis of multiple compounds, and safe and environmental protection, was thus used combined with stoichiometry to develop mathematical models, to assess those nutrient compounds. Therefore,the infrared spectra of 85 samples from P. ostreatusas fruiting bodies collected nationwide were determined. 5 kinds of spectral data pretreatment methods, multiple scatter correction (MSC), standard normal transformation (SNV), orthogonal signal correction (OSC), smooth plus first derivative (F-GD), and smooth plus second derivative (S-GD) were used. Following the model of the validation set regression coefficients, OSC combined with S-GD, and OSC combined with F-GD were the best pretreatment methods for the fruiting body protein and polysaccharide models. Under the optimal spectral pretreatment, 7 458 spectral bands were extracted by the LASSO algorithm, and 93 characteristic wavenumbers of protein and 92 for polysaccharides were obtained, with a compression rate of 98%. PLS model was established by fitting the characteristic wavenumbers with the protein and polysaccharide contents of P. ostreatus fruiting bodies detected by chemical method. The results showed that, for the protein model, the R2 regression coefficient of the calibration set was 0.999 8, RMSECV was 0.047 7, the R2 regression coefficientof the validation set was 0.987 2, RMSEP was 0.506 8, and RPD was 8.840 6 greater than 3, while for polysaccharides model, The R2 regression coefficient of calibration set was 0.999 9, RMSECV was 0.020 1, the R2 regression coefficient of validation set was 0.980 3, RMSEP was 0.292 9, and RPD was 7.119 8 greater than 3. The models thus had good predictive ability and robustness. This research provides a practical reference to determine a high-speed detection method for the nutrient content ofedible fungi by FTIR, a foundation to establish a nutritional quality evaluation for P.ostreatus and the promotion of their high-quality development, even for other edible fungi.