Differentiation in stingless bee honey composition
The pairwise comparison of SBH between each pair of tropical regions using PERMANOVA showed significant differences (p=0.003) in the composition of honeys across regions. However, the proportion of the variation explained by the geographic group was low (<20%, R² ranging from 0.05 to 0.183) , with the lowest value recorded being between the Neotropics and the Afrotropics (5%).
Random forests models and hierarchical clustering (Table 1 & Supplementary Fig. 3) illustrated that the distinction between groups of honey samples was ambiguous, particularly among samples from the Neotropics and the Afrotropics. None of the clusters included samples from a single tropical region; instead they were composed of a mix of samples from widely different geographical origins. Amino acids, fermentation makers and sugars were significantly varying among each tropical regions (Fig. 3), especially those involved in lactic acid fermentation (sugars including fructose glucose, raffinose and maltose, pyruvic acid and lactic acid) which were the most important compounds in the random forest classification. Indicator compound analysis showed that honeys from the Afrotropics were associated to kynurenic acid (stat = 0.424, p-value = 0.005) and acetoin (stat = 0.511, p-value = 0.01) and honeys from the Neotropics to methylglyoxal (stat = 0.532, p-value = 0.005) while we found no significant association between a specific compound and honeys from the Indo-Malayan region.
Table 1  : Prediction results of the conditional random forest models to predict the geographical origin of SBH samples.