Glycans are modifications added on to many mammalian proteins that can significantly affect their functions. They're very complicated though, and we don't have super good ways of predicting their structure. This study presents a dilated residual neural network approach to get glycan structure just from liquid-chromatography-MS/MS data (they trained their model on 500k annotated MS/MS spectra), which could help to push glycobiology forward and help us understand how glycan patterns contribute to protein function and disease.