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

Discarding low-quality merged MS/MS spectra. Even when considering all MS/MS spectra for some features, we sometimes have insucient information for both spectral library search and molecular formula annotation; to this end, such low-quality features were discarded. A feature is discarded if it produces less than 5 fragment peaks, estimated after merging peaks within 10 ppm or 0.0025 m/z from all corresponding MS/MS spectra; and if no fragmentation tree in the top 50 candidate list can explain at least 5 peaks accounting for at least 80 % of total spectrum intensity, see SIRIUS analysis below. Filtering low quality features decreased the number of features for dendroides from 1,078 to 784, for NIST1950 from 568 to 400, for tomato from 3,583 to 2,584, for diatoms from 3,227 to 2,075 and for mice stool from 577 to 377.