Biosynthetic glycan modeling reveals metabolic shifts in disease

Glycans are biosynthesized by partly overlapping cascades of enzymatic reactions that can be conceptualized as biosynthetic networks. These networks latently contain reaction rates, substrate specificities, and metabolic states of the underlying cells. Yet extracting this information, across tissues, species, and hundreds of datasets, is challenging with existing methods. Here, we present a new structure-aware algorithm for glycan biosynthetic network construction that is easily scalable to any number of glycomics datasets and results in more interpretable networks. We further show how these networks then can be used to elucidate substrate preferences of enzymes, quantify metabolic shifts in conditions, such as cancer, and provide hypotheses for not yet elucidated structures via physiological network extension. We also show that tumorigenesis shifts glycome regulation from a carrier-expression-dominated regime toward being governed by biosynthetic network structure, which generalizes across six independent cancer cohorts. We envision that these improvements to glycan biosynthetic modeling will aid in understanding glycan regulation and observed glycomic shifts in health and disease.