All raw data were centroided and converted to 32-bit uncompressed mzXML file using Bruker Data Analysis. A script was developed to select all possible MS/MS spectra in each LC-MS/MS run that could correspond to a compound present in the sample. For each compound, we calculated the theoretical mass M from its chemical composition and searched for the M+H, M+2H, M+K, and M+Na adducts. Putative identifications included all MS/MS spectra whose precursor m/z had a ppm error <50 compared to the theoretical mass of each possible precursor m/z; all tandem MS/MS spectra with an MS1 precursor intensity of <1E4 were ignored. All candidate identifications were manually inspected and the most abundant representative spectrum for each compound was added to the corresponding library at the gold or bronze level based upon an expert evaluation of the spectrum quality. The best MS/MS spectrum per compound as added to the GNPS-Collections library without filtering or alteration from the mzXML files.