Towards prediction of N-glycan compositions from atomic structural data

Glycobiology, the study of saccharides and their biological significance, delves into understanding glycans, oligosaccharides that form essential structures in various living organisms. However, these glycans, covalently linked to proteins or lipids, possess a structural complexity that exceeds that of nucleic acids and proteins, attributed to their non-templated assembly. This complexity, characterised by diverse linkage positions, degrees of branching, and isomerism, facilitates glycans' multifaceted roles, including cell-cell recognition, immune response, and protein function optimization.

Structural Biology is one of the fields concerned with the study of glycobiology, however current model-building software leans heavily towards proteins. A major hurdle is the absence of upfront knowledge of glycan compositions at glycosylation sites. While protein sequences are easily derived from DNA, glycan sequences are not directly encoded in genomes. As a result of these challenges, many modelled N-glycan chains in glycoproteins show errors as featured in numerous communications and remediation efforts. Therefore, part of the thesis was devoted to implementing a software solution that would enable scientists building atomic models of glycoproteins to easily access information retrieved from glycoproteomic studies. The new code, implemented as part of the Privateer carbohydrate model validation and analysis software, was demonstrated to be useful in validation of modelled N-glycan compositions during iterative model building.

Following the successful bridging of atomic coordinates and glycoproteomic data, the research pivoted to assess the interplay between amino acid identities and N-glycan composition. Limited data indicated a potential relationship, especially with aromatic amino acids. Thankfully, the advent of AlphaFold motivated the implementation of a grafting algorithm in the Privateer software, responsible for transplanting N-glycan atomic coordinates, therefore enabling the expansion of N-glycan atomic structure data. The development of new software tools enabled the discovery of potentially meaningful discriminatory relationships in terms of neighbouring amino acid chemical properties and the N-glycan processing products.