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

Hence, this dataset is particularly challenging for molecular formula assignment. Out of the 201 ground truth compounds, the preprocessing assigned an incorrect adduct to three; for these, the correct molecular formula is not contained in the candidate list considered by ZODIAC. For one compound, the corrected molecular formula was not ranked into the top 50. For the remaining 197 compounds, SIRIUS correctly annotated 50.76 % (100), compared to 96.95 % (191) for ZODIAC without anchors. This represents an 16.17-fold decrease in error rate. Error rates improves for compounds over the whole mass range, see Fig. 1 (right).