Skip to main content
CORESTA Meeting, Smoke Science/Product Technology, 2023, Cancun, STPOST 04

Quantitative analysis for metabolic profiling of aroma compounds by near infrared spectroscopy (NIR)

YANG Panpan(1,2); ZHOU Wenzhong(1,2); YANG Xiaoyun(1,2); LI Ming(1); LU Xiaoting(1,2); LIU Jing(1,2)
(1) Yunnan Reascend Tobacco Technology (Group) Co., Ltd, Kunming, China; (2) Yunnan Comtestor Co., Ltd, Kunming, China

It is well-known that the types and content of aroma components in tobacco is an important focus of the cigarette industry. Different types and contents of aroma components are the reaction of different metabolic profiling. This study proposed quantitative analysis for metabolic profiling of K326 tobacco aroma compounds’ PLS-DA model by NIR, which had been planted in three typical ecological environments (Henan, Guizhou and Yunnan Provinces of China). Firstly, the GC-MS method was used to detect the kinds and content of aroma components of K326 flue-cured tobacco. A partial least squares discriminant analysis (PLS-DA) model was built to develop the feature parameters, the principal component score, of the metabolic profiling of the aroma components. Secondly, using the partial least squares (PLS) method, the model between the near infrared spectrum data and aroma components of metabolic features of relevant parameters was fitted. Thirdly, the internal parameters that include the correlation coefficient (R2), the appropriate number of principal components (k), the standard error of calibration (SEC) and root mean square error of cross validation (RMSECV) were calculated to evaluate the models; and the external parameters that include average error verification, standard error of validation (SEV), standard deviation of validation error (SDV), the t distribution of paired t test value and probability P were also calculated to evaluate the models. Results show that (1) The PLS-DA model can clearly develop the characteristics of metabolic profiling of aroma compounds; (2) The score of component 1 correction model of internal evaluation parameters: R2, k, SEC and RMSECV is 0.970, 9, 0.829 and 0.976, respectively; the external evaluation parameters: average error verification, SEV, SDV, the t distribution of paired t test value and probability P is 0.00291, 0.09960, 0.10083, 0.09960 and 0.856, respectively. The score of component 2 correction model of internal evaluation parameters: R2, k, SEC and is 0.908, 7, 1.13 and 1.22, respectively; the external evaluation parameters: average error verification, SEV, SDV, the t distribution of paired t test value and probability P is 0.02111, 0.12606, 0.12586, 0.12606 and 0.295, respectively. Thus, the established model is of good stability and prediction accuracy. The establishment of the method for using near infrared spectroscopy technology combined with chemometrics to rapidly quantitatively predict metabolic profiling of K326 tobacco aroma compounds is a good and significant practice.