Abstract
High-Performance Thin-Layer Chromatography (HPTLC) was used in a chemometric investigation of the derived sugar and organic extract profiles of two different honeys (Manuka and Jarrah) with adulterants. Each honey was adulterated with one of six different sugar syrups (rice, corn, golden, treacle, glucose and maple syrups) in five different concentrations (10%, 20%, 30%, 40%, and 50% w/w). The chemometric analysis was based on the combined sugar and organic extract profiles' datasets. To obtain the respective sugar profiles, the amount of fructose, glucose, maltose, and sucrose present in the honey was quantified and for the organic extract profile, the honey's dichloromethane extract was investigated at 254 and 366 nm, as well as at T (Transmittance) white light and at 366 nm after derivatisation. The presence of sugar syrups, even at a concentration of only 10%, significantly influenced the honeys' sugar and organic extract profiles and multivariate data analysis of these profiles, in particular cluster analysis (CA), principal component analysis (PCA), principal component regression (PCR), partial least-squares regression (PLSR) and Machine Learning using an artificial neural network (ANN), were able to detect post-harvest syrup adulterations and to discriminate between neat and adulterated honey samples. Cluster analysis and principal component analysis, for instance, could easily differentiate between neat and adulterated honeys through the use of CA or PCA plots. In particular the presence of excess amounts of maltose and sucrose allowed for the detection of sugar adulterants and adulterated honeys by HPTLC-multivariate data analysis. Partial least-squares regression and artificial neural networking were employed, with augmented datasets, to develop optimal calibration for the adulterated honeys and to predict those accurately, which suggests a good predictive capacity of the developed model.