Deep learning has shown significant success in mass spectrometry-based proteomics and is now being applied to glycoproteomics. However, existing deep learning models struggle with the non-linear glycan structure in intact glycopeptides. To address this, the authors present DeepGlyco, a deep learning-based approach for predicting fragment spectra of intact glycopeptides. The model uses tree-structured long-short term memory networks to process the glycan moiety and graph neural networks to incorporate potential fragmentation pathways of specific glycan structures. This approach enhances model explainability and the differentiation of glycan structural isomers. The predicted spectral libraries are shown to be useful for data-independent acquisition glycoproteomics, improving library completeness. The study demonstrates that DeepGlyco can accurately predict MS/MS spectra of intact glycopeptides, with high performance across different organisms and instrument settings. The model's ability to differentiate glycan structural isomers and its potential for enhancing DIA data analysis are highlighted. The work provides a valuable resource for the glycoproteomics community, with potential applications in other informatic workflows.Deep learning has shown significant success in mass spectrometry-based proteomics and is now being applied to glycoproteomics. However, existing deep learning models struggle with the non-linear glycan structure in intact glycopeptides. To address this, the authors present DeepGlyco, a deep learning-based approach for predicting fragment spectra of intact glycopeptides. The model uses tree-structured long-short term memory networks to process the glycan moiety and graph neural networks to incorporate potential fragmentation pathways of specific glycan structures. This approach enhances model explainability and the differentiation of glycan structural isomers. The predicted spectral libraries are shown to be useful for data-independent acquisition glycoproteomics, improving library completeness. The study demonstrates that DeepGlyco can accurately predict MS/MS spectra of intact glycopeptides, with high performance across different organisms and instrument settings. The model's ability to differentiate glycan structural isomers and its potential for enhancing DIA data analysis are highlighted. The work provides a valuable resource for the glycoproteomics community, with potential applications in other informatic workflows.