ZODIAC: database-independent molecular formula annotation using Gibbs sampling reveals unknown small molecules

Marcus Ludwig; Louis-Félix Nothias; Kai Dührkop; Irina Koester; Markus Fleischauer; Martin A. Hoffmann; Daniel Petras; Fernando Vargas; Mustafa Morsy; Lihini Aluwihare; Pieter C. Dorrestein; Sebastian Böcker

We use strict lters for both the ZODIAC score, the quality of the underlying MS/MS data, and the support by other molecular formulas in the dataset. We report molecular formula annotations from all ve datasets with a) minimum ZODIAC score of 0.999, b) at least 95 % of the MS/MS spectrum intensity being explained by SIRIUS, and c) at least one molecular formula of the compound is connected to 20 or more compounds in the ZODIAC similarity network. The third criterion discards compounds where ZODIAC's results are basically identical to SIRIUS's. This results in three novel molecular formulas in the tomato dataset and 59 in the diatoms dataset, see Supplementary Table 3 and 4. Filtering less restrictively (ZODIAC score at least 0.99, at least 90 % of the MS/MS spectrum intensity being explained by SIRIUS), we annotate 31 in tomato, 103 novel molecular formulas in the diatoms dataset, one in NIST1950 and one in mice stool.