Restoring Protein Glycosylation with GlycoShape

Abstract

AbstractDuring the past few years, we have been witnessing a revolution in structural biology. Leveraging on technological and computational advances, scientists can now resolve biomolecular structures at the atomistic level of detail by cryogenic electron microscopy (cryo-EM) and predict 3D structures from sequence alone by machine learning (ML). One technique often supports the other to provide the view of atoms in molecules required to capture the function of molecular machines. An example of the extraordinary impact of these advances on scientific discovery and on public health is given by how structural information supported the rapid development of COVID-19 vaccines based on the SARS-CoV-2 spike (S) glycoprotein. Yet, none of these new technologies can capture the details of the dense coat of glycans covering S, which is responsible for its natural, biologically active structure and function and ultimately for viral evasion. Indeed, glycosylation, the most abundant post-translational modification of proteins, is largely invisible through experimental structural biology and in turn it cannot be reproduced by ML, because of the lack of data to learn from. Molecular simulations through high-performance computing (HPC) can fill this crucial information gap, yet the computational resources, the users’ skills and the long timescales involved limit applications of molecular modelling to single study cases. To broaden access to structural information on glycans, here we introduce GlycoShape (https://glycoshape.org) an open access (OA) glycan structure database and toolbox designed to restore glycoproteins to their native functional form by supplementing the structural information available on proteins in public repositories, such as the RCSB PDB (www.rcsb.org) and AlphaFold Protein Structure Database (https://alphafold.ebi.ac.uk/), with the missing glycans derived from over 1 ms of cumulative sampling from molecular dynamics (MD) simulations. The GlycoShape Glycan Database (GDB) currently counts over 435 unique glycans principally covering the human glycome and with additional structures, fragments, and epitopes from other eukaryotic and prokaryotic organisms. The GDB feeds into Re-Glyco, a bespoke algorithm in GlycoShape designed to rapidly restore the natural glycosylation to protein 3D structures and to predictN-glycosylation occupancy, where unknown. Ultimately, integration of GlycoShape with other OA protein structure databases can provide a step-change in scientific discovery, from the structural and functional characterization of the active form of biomolecules, all the way down to pharmacological applications and drug discovery.