Abstract
In the honey industry it is very important to define the distinctive characteristics of unifloral honeys and to find efficient methods for their verification which allow them to be competitive in the market. The present work was undertaken on different types of New Zealand unifloral honeys and honeydew honey to characterize the honey floral origin and to differentiate various types of honey using five different approaches including physicochemical analysis, investigation of rheological and thermal behaviour of honeys, determination of phenolic compounds and antioxidant activities, evaluation of plant-derived toxic compounds as markers and volatile analysis. The most discriminating parameters were selected from each analytical approach and were applied to discriminate the pollen-identical manuka and kanuka honeys.
The results of physicochemical analysis were subjected to principal component analysis (PCA), artificial neural networks (ANN) and a combination of PCA and ANN. PCA explained 64% of the variations between sample types. In order to find the best ANN model with the least number of parameters for distinguishing the honey types, a stepwise elimination technique was applied to the data. The optimum model with the most efficient classification and the least number of parameters was achieved with 5 parameters including conductivity, colour, glucose, fructose and sucrose levels. The combination of PCA and ANN provided the most efficient discriminatory model with the accuracy of 0.92 and regression coefficient of 0.96.
The thermal and rheological analysis demonstrated that rheology values, glass transition temperatures and thermal decomposition events were influenced by the origin of honey samples. Discriminant analysis was conducted to determine the discriminating power of thermal and rheological properties of honey and to determine the most effective parameters for differentiation of honey samples according to their floral origin. The first discriminant function accounted for 78.7% of total variance while the second accounted for 13.7%. Samples of thyme, honeydew and manuka were clearly differentiated from other samples. The most powerful classifying parameters were viscosity at 10˚C, 0˚C, and 20˚C according to linear discriminant analysis.
Total phenolic content (TPC), antioxidant activity (DPPH and FRAP assays), colour and the profile of flavonoids and phenolic compounds were also affected by honey origin. A linear discriminant analysis (LDA) using the 10 most discriminating parameters successfully discriminated the honey types; explaining 73% variation by the first two functions. In the LDA discriminant model, seven variables were eliminated and ten variables remained (syringic acid, quercetin, benzoic acid, kaempferol, chlorogenic acid, colour, p-coumaric acid, myricetin, gallic acid and FRAP).
The incidence of echimidine as a specific floral marker of Echium vulgare spp. was evaluated in New Zealand honey samples. The presence of echimidine, a major hepatotoxic pyrrolizidine alkaloid produced by E. vulgar was significantly present in 67% of viper’s bugloss honey samples in the range of 2.6-52. µg.kg-1. However, neither the floral samples nor the honeydew honey contained echimidine at detectable levels.
One hundred and sixteen volatile compounds were identified in 29 honey samples by analysis of volatile compounds. In order for differentiation of the samples, PCA and partial least square discriminant analysis (PLS-DA) were performed. The first three components of PCA accounted for 57.3% of variation and indicated differentiation between honey types with some overlaps between confidence intervals of the honey clusters. PLS-DA demonstrated total variance of 45.1% using components 1 to 3 and showed clear separation between honey samples. Results suggested that the analysis of volatile compounds followed by multivariate data analysis was the most efficient method for differentiation of the honey samples according to their origin.
As manuka and kanuka honeys are indistinguishable by melissopalynology, selected discriminating parameters, including conductivity, pH, colour and FRAP values, rheology analysis and volatile analysis were applied to attempt discrimination of manuka from kanuka honey. Using physicochemical parameters, FRAP and viscosity values were not successful in discriminating and clustering of manuka and kanuka honey samples. However, the presence and level of volatile constituents showed high discriminating capacity between the pollen-identical manuka and kanuka honey samples.