Characterization of volatile flavor profiles and quantitative assessment of key physicochemical indicators for fermented bean curd using GC-IMS, E-nose and multi-channel colorimetric sensor array combined with chemometrics

To elucidate the volatile flavor compositions and overall aroma profiles of fermented bean curd (FBC), the volatile organic compounds (VOCs) in six different FBC were detected and characterized using gas chromatography-ion mobility spectrometry (GC-IMS) and an electronic nose (E-nose) in this study. A total of 60 VOCs were identified by GC-IMS, of which esters, aldehydes, alcohols, and ketones constituted the major compounds. Among them, 17 VOCs were identified as key differentiating volatile compounds. In addition, the E-nose combined two algorithms, linear discriminant analysis (LDA) and k-nearest neighbor (KNN), to demonstrate its effectiveness in differentiating between different FBC samples. The results showed that the LDA model performed better than the KNN model. When the principal component number was 9, the recognition accuracies of the training and prediction sets for the LDA model were 94.44 % and 91.67 %, respectively. In addition, a multi-channel colorimetric sensor array (CSA) was constructed in this study for the quantitative prediction of key physicochemical indicators. The results showed that both partial least squares regression (PLSR) and support vector machine regression (SVR) achieved good prediction performance. Among them, for the SVR model, the prediction correlation coefficients for total acidity, reducing sugar, salinity, and amino acid nitrogen were 0.9033, 0.9170, 0.7298, and 0.9213, respectively. The results of this study indicate that GC-IMS, E-nose, and CSA are expected to be effective tools for characterizing FBC flavor as well as facilitating the rapid quantification of key physicochemical indicators, which may provide valuable insights for flavor and quality control in traditional fermented foods.