Integration of in silico Annotation, Automated Chemical Classification, and Substructure Recognition With Mass Spectral Molecular Networks

In silico structure annotation was performed by submitting the preprocessed ms1-2.mgf output file of the pooled extracts from Optimus to Sirius+CSI:FingerID (Dührkop et al., 2015; Böcker and Dührkop, 2016) with m/z tolerance set to 20 ppm. Additionally, data were submitted to Network Annotation Propagation (NAP) (da Silva et al., 2018). For NAP, both [M+NH4]+ and [M+H]+ adducts were searched with m/z tolerance set to 15 ppm and parameters described at: https://proteomics.ucsd.edu/ProteoSAFe/status.jsp?task=184a80db74334668ae1d0c0f852cb77c and https://proteomics2.ucsd.edu/ProteoSAFe/status.jsp?task=2cfddd3b8b1e469181a13e7d3a867a6f.

We matched a custom database of molecular structures against our samples’ preprocessed mass spectral data using both Sirius+CSI:FingerID and NAP. This database was compiled manually from literature (Shi et al., 2008; Vasas and Hohmann, 2014) and the dictionary of natural products (DNP)³. Subsequently, in silico structure matches from Sirius+CSI:FingerID and NAP were submitted to automated chemical classification using ClassyFire⁴ (Djoumbou Feunang et al., 2016) and consensus classifications at each hierarchical level of the chemical taxonomy per mass spectral molecular subnetwork were calculated. Consensus classifications and molecular structures were then visualized on the molecular networks using Cytoscape version 3.4.0 (Shannon et al., 2003). Substructure recognition of the crude extracts was performed by submitting the preprocessed ms2.mgf output file from Optimus to MS2LDA (van der Hooft et al., 2016; Wandy et al., 2018). Data and parameters used are publically accessible at http://ms2lda.org/basicviz/short_summary/390/. Subsequently substructures (Mass2Motifs) were mapped on the nodes of the mass spectral molecular network, and Mass2Motifs shared among different nodes were mapped on the edges connecting the nodes and visualized using Cytoscape version 3.4.0. Description of the integration of in silico annotation, automated chemical classification, and substructure recognition with mass spectral molecular networks for 3D mass spectral molecular cartography is provided in the Supplementary Methods.