Understanding glycan structure and function through artificial intelligence

Glycoinformatics has entered the artificial intelligence era. Stymied by a lack of big data, high sequence complexity, and biosynthetic dependencies, the application of machine learning to glycomics data has largely emerged this decade. In this mini-review, we explore the latest groundbreaking computational approaches applied to glycan sequencing, classifying disease risk from glycan biomarkers, and predicting protein–glycan interactions. We detail advancements in the architectures of these models, concluding that they have matured to the stage of extracting predictive information from glycan sequences. We also discuss their challenges and limitations, and how to fully reap their rewards in the future.