Perceptual Interactions Of Aroma Presentation And Construction Of Odor Intensity Prediction Model Of Wine Terpenoids

Terpenes are an important component of wine varietal aromas,with low odor threshold and elegant aroma profile.At present,only the threshold values of a few terpenes are known,and there is a diversity of relevant reference values.On the other hand,the interactions between compounds have a significant impact on the overall aroma perception of wine,however,the interactions between terpenes have been poorly studied.Therefore,the analysis of the odor characteristics of terpenes in grapes and wines and the investigation of their interactions are important theoretical guidance for the regulation of wine aroma quality.In this paper,we analyzed the odor characteristics of 28 typical terpenes in grapes and wines,including the determination of odor descriptors and threshold,and explored the interactions between terpenes,and finally developed a prediction model to predict the odor intensity of mixtures based on compound concentrations and thresholds.The main research results are as follows:（1）Odor description and odor threshold of terpenes.The results showed that there is a significant influence of structure on the odor description and threshold of terpenes.According to the structure and aroma,terpenes can be divided into three categories:linear terpenes（floral and fruity）,monocyclic monoterpenes（herbal and spicy）,epoxy monoterpenes（rose）.In this study,the lowest threshold was rose oxide（2.7μg/L）,and the largest threshold wasβ-citronellene（4005.5μg/L）.The matrix effect on terpenes varied,and in general,the presence of wine matrix increased the threshold of compounds,with dry white matrices generally having a weaker effect on compounds than dry red matrices.（2）Perceptual interactions among terpenes.The results showed that additive or synergistic effect is common between terpenes,and masking or compromising interactions exist between some compounds.Among them,binary mixture of terpenes with the same functional group is generally additive or synergistic,"nerol+geraniol","citral+citronellal","linalyl acetate+geranyl acetate"and so on are additive effects,"geraniol+linalool","β-myrcene+citronellene","rose oxide+linalool oxide","terpine-4-ol+α-terpineol"and so on are synergistic effects.While the interaction effect between compounds with similar structure or aroma and other compounds differ significantly,for example,the interaction effect between geraniol and other terpenes is generally stronger than that between linalool,and farnesol is generally stronger than nerolidol.When multiple terpene binary mixtures with the same effect are formed a multiple mixture may have the same effect or may produce stronger interactions.The results ofσ-τdiagram show that partial addition occurs mainly between terpenes of different concentration levels,and additive or synergistic interactions is common among terpene compounds at suprathreshold concentrations,while the interactions between terpene compounds at subthreshold concentrations depend on the type of compound.（3）Odor intensity prediction model of terpenes.The results showed that the odor intensity of terpenes and the logarithm of their odor activity values conformed to the linear equation OI=1.42ln OAV.The total intensity of the terpene binary mixture and the sum of the odor intensity of single compounds conformed to the linear equation OI_Mea=0.61OI_Sum.Based on this,the additivity model,the U model and the vector model are established,in which the U model and the vector model predicted the mixture intensity significantly better than the additivity model.The PDE model was also established to explore the effect of mixing ratio on the odor intensity of the mixture,and to predict the mixture intensity.The results showed that the PDE model has good prediction accuracy,and the prediction effect is better than the additivity model,the U model and the vector model.In addition,the extended PDE model of terpene mixture（y=2.38x²-2.38x+1.53）was obtained by summing up and fitting all the PDE models.The accuracy verification test proved that the extended PDE model also had good prediction effect.