Mathematical Model Decodes How IgG Glycan Patterns Change With Age

IgG glycosylation—the attachment of sugar chains to antibodies—is not static. It shifts with age, disease, and genetics, and these shifts carry diagnostic and functional consequences. Yet translating raw glycan measurements into mechanistic insight has remained difficult. This study takes a key step forward by building and validating a quantitative mathematical model that reconstructs the enzymatic activity underlying IgG glycan profiles from large-scale population data.

The researchers constructed a rule-based kinetic model of IgG N-glycosylation spanning four Golgi compartments (cis-, medial-, trans-cisternae, and the trans-Golgi network). The model begins with the M5 mannose precursor and simulates the sequential actions of seven enzymes: GnT I, GnT II, GnT III, Man II, FucT, GalT, and SiaT. It was calibrated using UHPLC-measured glycan data from 915 individuals on Korčula Island and validated in an independent cohort of 890 individuals from Vis Island, both in Croatia. Each individual was represented by 22 chromatographic glycan peaks, and the model was personalized by fitting six individual enzyme concentration parameters per person.

The model achieved strong agreement with experimental glycan distributions in both cohorts. Crucially, it revealed a consistent, statistically significant age-related decline in GalT concentration across both populations. GalT is responsible for adding galactose to terminal GlcNAc residues—a well-known step that diminishes with aging and drives the shift toward agalactosylated glycan forms linked to inflammaging. The model effectively recovered this biological signal from glycan profiles alone, without direct enzyme measurement.

Beyond recapitulating known biology, the model demonstrated that recovered GalT and other enzyme concentrations could predict chronological age with accuracy comparable to established glycan peak-based aging biomarkers. This positions the modeled enzyme activities as a mechanistically interpretable layer between genetic variants and glycan outputs—potentially making genome-wide association studies (GWAS) more powerful and biologically interpretable by targeting enzyme-level variation rather than individual glycan peaks.

The authors envision this as the first step in a two-stage strategy: recover enzyme activities via mathematical modeling, then map genetic variants to those activities. This could improve detection power for SNPs whose effects are diluted or obscured when analyzed against complex glycan peak composites. Limitations include the exclusion of early ER and cis-Golgi processing, reliance on populations of European ancestry, and the use of relative rather than absolute glycan concentrations. Nevertheless, the approach is computationally tractable, openly available via the BioUML platform, and extensible to other glycoprotein systems.