Fig. 1.

Feature-based molecular network displaying the 15 predominant putative chemical classes and their subclasses. Nodes represent mass spectral features and are used as a proxy for a metabolite. Connected nodes represent high tandem mass spectral similarity, and thus high chemical structural similarity. The thickness of the gray edges connecting nodes varies according to the cosine score representing to what extent two connected metabolites are chemically similar (based on MS2 spectra, from 0.7: less similar and thin edge to 1.0: identical and thick edge). The name of annotated metabolites (levels 1 and 2), details on chemical classes with fewer than 4 metabolites (absent on this figure), chemical classification scores (Ernst et al. 2019), all unknowns, and group intensities for all features (average, standard deviations) are detailed in Online Resource 6. Note that data represent a summary of most predominant classes per molecular family retrieved through either GNPS spectral library matching or in silico structure annotation and may contain false positives