Discrimination of different stir-frying degrees of Hordei Fructus Germinatus based on GC-IMS and machine learning

Abstract

Hordei Fructus Germinatus (MY) is a functional food primarily existing in two forms, namely sheng MY and chao MY, after stir-frying process. At present, the identification of stir-frying degree of MY predominantly relies on the subjective judgement of pharmaceutical scientists, which lacks clear, quantifiable indicators. In this study, Gas Chromatography-Ion Mobility Spectrometry (GC-IMS) was employed to identify the volatile organic components (VOCs) in MY with different stir-frying time, a total of 54 VOCs were identified. Subsequently, a total of eight featured VOCs were screened, including (E)-2-hexanal, (Z)-3-hexenyl propionate, (2S-trans)-cyclohexanone, (E)-2-pentenal, ethyl 2-methylbutanoate, 3-methylpentanoic acid, ethyl formate, and methyl acetate. These changes in featured VOCs content were attributed to chemical reactions, including oxidation reaction, degradation reaction, and Maillard reaction. Finally, based on the featured VOCs, seven machine learning models were established, and their performance was comparatively studied. The results show that the three models, SVM-L, SVM-P and NB, could accurately discriminate MY with different stir-frying time, and the accuracy of all three models was above 96% under both training and testing sets. Overall, this study presents a rapid, low-cost, and accurate method for discriminating MY samples with different stir-frying degrees, providing a theoretical foundation for the quality control of stir-frying of MY processes.