Molecular modelling of glycoproteins by homology with non-glycosylated protein domains, computer simulated glycosylation and molecular dynamics.

OBJECTIVES: This study aims to visualise glycoproteins by computer graphics molecular modelling in order to research the dynamics of the oligosaccharide chains, determine their affects on protein conformation, antigenicity and function and to characterise oligosaccharide recognition determinants. With respect to the last, the modelling included the sialylpolylactosamine of thymocyte Thy-l and the sialyl Le(x)/Le(a) determinant present on brain Thy-l.
METHODS: The following techniques were used: 1) database searching for homologies with non-glycosylated protein domains; 2) protein modelling on the basis of homology and secondary structure prediction techniques; 3) oligosaccharide construction using a simulated annealing approach utilising the AMBER forcefield with appropriate parameters in the Biosyn software environment; 4) creation of glycoprotein conjugates for further investigation by energy minimisation and molecular dynamics.
RESULTS: This approach was successful in providing models of Thy-l and the carboxy terminus 27.5 kD of HIV-1 gp 120, by homology with immunoglobulin light chain folds and in one case (Thy-l) adding the oligosaccharide chains, phosphatidylinositol glycan anchor and lipid membrane and, in the other, adding additional highly glycosylated domains, and domains which folded by molecular dynamics. Significant affects on protein conformation were shown in the presence or absence of the lipid anchor and simulated membrane, but not by the N-linked oligosaccharide chains.
CONCLUSIONS: The highly glycosylated molecules Thy-l and gp120, which are not expected to crystallise in their native state, were modelled by computer graphics simulated annealing or molecular dynamics from which interactions could be predicted which agree with experimental data on antibody binding and in vitro activity.