Prediction of pyrazines and identification of flavor intensity in boletus bainiugan at different drying temperatures...

Low-cost and high-throughput quantitative analysis of volatile compounds is essential for flavor studies in Boletus bainiugan. However, traditional methods are time-consuming and labor-intensive. Herein, a rapid prediction method of pyrazines based on spectroscopic and data fusion techniques for Boletus bainiugan of different drying temperatures is proposed. Four datasets, Fourier transform Near infrared (FT-NIR) spectroscopy, Attenuated total reflectance Fourier transform infrared (ATR-FTIR) spectroscopy, and low-level and middle-level data fusion, are used to build 147 partial least squares regression (PLSR) prediction models based on different preprocessing and feature extraction methods. The results indicates that the second-order derivation- uninformative variable elimination (SD-UVE) based FT-NIR spectral dataset achieved accurate prediction for seven important compounds (pyrazines), with the coefficient of determination of prediction set (R²_p) higher than 0.840 and residual predictive deviation (RPD) are all greater than 2.418. The screened 189 feature wavenumbers can be used to accurately predict other pyrazine compounds with the highest R²_p of 0.909 and the best RPD of 2.506, and successfully differentiate between Boletus bainiugan with high flavor intensity (65 °C) and other drying temperature samples with a validation set accuracy of 95.23 %. This study achieved the work of predicting the relative content of pyrazine volatile compounds in different drying temperatures of Boletus bainiugan, which provides a theoretical basis for the study of other classes of compounds in spectral distribution and prediction.