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

It is understood that some of these annotations may be wrong; unfortunately, a complete evaluation would require a full structural elucidation, which is experimentally infeasible. But our results clearly show that ZODIAC allows the user to select a few, potentially highly interesting compounds from a set of hundreds or thousands with low eort. Furthermore, using an example, we show in the next section that one top-scoring annotation from Supplementary Table 3 is presumably correct. Detailed evaluation of a novel bromine-containing molecular formula. We now concentrate on one particular compound in the diatoms dataset (m/z 588.230, retention time 503.97 sec): This ZODIAC-annotated compound is protonated and has molecular formula C 24 H 47 BrNO 8 P, which is indeed absent from the structure databases. The occurrence of bromine agrees with our expectation that marine organisms can be prolic sources of organohalogens 36 . The ZODIAC score of this annotation is 1.0, the maximum value. We found multiple lines of evidence that this molecular formula annotation is correct, both in the measured isotope pattern and the three MS/MS spectra measured for m/z 588.230 (presumably the monoisotopic peak of the isotope pattern), 590.228 (presumably the M+2 peak) and 592.325 (presumably the M+4 peak), see Fig. 3 . To annotate fragments with molecular formulas, we used SIRIUS to compute a fragmentation tree for the MS/MS spectrum of the monoisotopic peak at m/z 588.230 (Fig. 3e ).