Metabolite correlation-based network analysis combined with machine learning techniques highlights LOX biosynthesis in Vanilla planifolia and Vanilla pompona source leaves

Abstract The vanilla genus of the orchid family is the primary source of the famous vanilla spice. Here, the central metabolome of source leaves of two important commercial vanilla species (V. planifolia, V. pompona) - collected from two distinct geographical locations -  was profiled using a GC-TOF–MS platform. In total, 544 metabolites were identified which were subjected to a combination of multivariate and univariate analysis, location-adjusted models, correlation-based network analysis (CNA), and CNA combined with machine-learning assisted pathway mapping. Multivariate analysis of the metabolic profiles revealed a clear separation between the two species, confirmed by location-adjusted models. Univariate statistical analysis highlighted linoleic acid in the V. planifolia vs. V. pompona comparison. CNA showed higher connectivity in the V. pompona network over the V. planifolia network (6425 vs. 3508 edges), suggestive for biochemical adaptations for each species. CNA combined with machine learning techniques highlighted the lipoxygenase (LOX) pathway. This finding, combined with the identification of linoleic acid during univariate statistical analysis, indicated modified fatty acid metabolism in V. pompona with potential consequences for attracting pollinators.