基于游离氨基酸的烤烟配打模块香型判别方法研究

Identification Methods for Flue-cured Tobacco Flavor in Blending and Threshing Module Based on Free Amino Acids

为明确不同香型烤烟配打模块游离氨基酸组分差异,并实现对清香型、中间香型及浓香型烤烟复烤模块的准确分类,选取7个清香型、7个中间香型和8个浓香型烤烟配打模块为研究对象,对其游离氨基酸含量进行OPLS-DA分析,结合SVM模型进行香型判别。结果表明:不同香型烤烟模块游离氨基酸总量范围为13141.72~27695.47μg/g,其中脯氨酸和天冬酰胺含量在各香型中占比最高,天冬氨酸、谷氨酸、甘氨酸、丙氨酸、胱氨酸、亮氨酸、酪氨酸、苯丙氨酸、4-氨基丁酸、赖氨酸、组氨酸、色氨酸、精氨酸和游离氨基酸总量在不同香型间存在显著性差异。丙氨酸和天冬氨酸在OPLS-DA分析中VIP值大于1且在不同香型间差异显著(p<0.05),可筛选为香型判定指标。基于21种游离氨基酸的前两种主成分构建的SVM模型对模块香型的判别率准确率为86.36%;基于筛选出的2个游离氨基酸特征指标构建的SVM模型判别准确率为90.91%。基于丙氨酸和天冬氨酸的SVM模型判别率和可信度都在较高水平,可应用于烤烟配打模块的香型判别。

Aiming to clarify the differences of free amino acid components in different flavor types of flue-cured tobacco blending and threshing modules,and classify redrying modules of fresh flavor,neutral flavor and robust flavor types,the blending modules of 7 fresh,7 neutral and 8 robust flavor types were selected,and their free amino acid contents were determined by OPLS-DA analysis,flavor types were identified by SVM model.Results showed that total free amino acids in different flavor types of flue-cured tobacco ranged from 13141.72 to 27695.47μg/g,and the contents of proline and asparagine exhibited the highest among all flavor types.Significant differences among different flavor types occurred in the total amount of aspartate,glutamic acid,glycine,alanine,cystine,leucine,tyrosine,phenylalanine,4-aminobutyric acid,lysine,histidine,tryptophan,arginine and free amino acid.In OPLS-DA analysis the VIP values of alanine and aspartic acid were above 1 and significantly different among various flavor types(p<0.05),hence could be selected as evaluation indexes.The SVM model based on the first two principal components of 21 free amino acids exhibited an accuracy of 86.36%.The discriminant accuracy of SVM model based on two free amino acid characteristic indexes reached 90.91%.The SVM model based on alanine and aspartic acid exhibited high discrimination rate and reliability,and can be applied in the flavor discrimination of flue-cured tobacco blending modules.