Green and rapid quantitative detection of selenium in selenium-enriched kefir grain based on Fourier transform infrared spectroscopy

Rapid monitoring of total and organic selenium content in kefir grain was essential for microbial screening and selenium-enriched food development. Firstly, spectral information of selenium-enriched kefir grain was obtained using an attenuated total reflection Fourier transform infrared spectrometer. Secondly, the performance of the quantitative prediction models established by the four-variable screening method with three machine learning algorithms, respectively, was compared. For the prediction of total selenium, the competitive adaptive reweighted sampling - least squares support vector machine model performed the best, with prediction set relative coefficient (R_P) and relative prediction deviation (RPD) values of 0.97 and 4.36, respectively. For the prediction of organic selenium, the IRF-LSSVM model had a R_P and RPD value of 0.95 and 6.44, respectively. The proposed method achieves scientific, rapid (within 1 min) and green detection of total selenium (237.72-2330.82 μg/g) and organic selenium (102.20-1483.59 μg/g) content in selenium-enriched Kefir grain.