Abstract: In order to improve the defects of the method specified in the Chinese national standard to determine the total sugar content in edible fungi (Lentinula edodes), such as long detection time and tedious steps, near infrared spectroscopy (NIR) combined with partial least square (PLS) was adopted to establish a rapid method for determining the total sugar content of L. edodes and the predictive model parameters were optimized. In this study, a total of 106 samples were collected, 13 of which were randomly assigned into the validation set for testing model reliability, while the remaining 93 were assigned into the calibration set. In the calibration set, according to the studentized residuals and leverage value, 65 lots of samples with the typical characteristics of L. edodes were selected to optimize spectral range and the number of PLS factors for the establishment of a quantitative model for predicting the total sugar content in L. edodes. It was found that the optimal spectral preprocessing method was multiplicative scatter correction (MSC) combined with second derivatives (SD), and the predictive model with optimal results was constructed using the spectral range of 4 000–10 000 cm-1 and 10 PLS factors, with a root mean square error of calibration (RMSEC) of 1.393 and a root mean square error of prediction (RMSEP) of 1.557 as well as a correlation coefficient for calibration (R2) up to 0.940 04 and a relative percent deviation (RPD) up to 4.08. Finally, the verification results showed the model-predicted value had a good linear relationship with the experimental value with no significant difference between them (P = 0.993). In conclusion, the NIR prediction model could accurately predict the total sugar content of L. edodes.

Key words: Lentinula edodes; near infrared spectroscopy; total sugar; partial least square; rapid detection method