Fig. 2: Discovering diagnostic fragmentation using CandyCrumbs.
From: Predicting glycan structure from tandem mass spectrometry via deep learning
a, Schematic view of the CandyCrumbs workflow for automatic fragment ion annotation. b–e, Negative ion mode spectra of reduced glycans with prediction confidence between 0.9 and 1.0 for Fucα1-2Galβ1-3GalNAc/Galβ1-4GlcNAcβ1-3Fuc (b), Neu5Acα2-3Galβ1-3GalNAc/Galβ1-3(Neu5Acα2-6)GalNAc (c), GlcNAcβ1-3(Neu5Acα2-6)GalNAc/GalNAcα1-3(Neu5Acα2-6)GalNAc (d) and GlcNAcβ1-3(Neu5Gcα2-6)GalNAc/GalNAcα1-3(Neu5Gcα2-6)GalNAc (e) were averaged and juxtaposed. Fragments exhibiting differential abundance were labeled by CandyCrumbs in the Domon–Costello nomenclature30. f,g, Negative ion mode spectra of reduced glycans with prediction confidence between 0.6 and 1.0 for Neu5Acα2-3Galβ1-4GlcNAcβ1-2Manα1-3(Neu5Acα2-6Galβ1-4GlcNAcβ1-2Manα1-6)Manβ1-4GlcNAcβ1-4GlcNAc/Neu5Acα2-3Galβ1-4GlcNAcβ1-2Manα1-3(Neu5Acα2-6Galβ1-4GlcNAcβ1-2Manα1-6)Manβ1-4GlcNAcβ1-4GlcNAc (f) and Neu5Acα2-6Galβ1-4GlcNAcβ1-2Manα1-3(Galβ1-4GlcNAcβ1-2Manα1-6)Manβ1-4GlcNAcβ1-4GlcNAc/Neu5Acα2-3Galβ1-4GlcNAcβ1-2Manα1-3(Galβ1-4GlcNAcβ1-2Manα1-6)Manβ1-4GlcNAcβ1-4GlcNAc (g) were averaged, juxtaposed and labeled similar to b–e. Doubly charged fragment ions are colored gray.
