• A novel graph-based method for targeted ligand-protein fitting

      Hannaford, Gareth James (University of Bedfordshire, 2008-08)
      The determination of protein binding sites and ligand -protein fitting are key to understanding the functionality of proteins, from revealing which ligand classes can bind or the optimal ligand for a given protein, such as protein/ drug interactions. There is a need for novel generic computational approaches for representation of protein-ligand interactions and the subsequent prediction of hitherto unknown interactions in proteins where the ligand binding sites are experimentally uncharacterised. The TMSite algorithms read in existing PDB structural data and isolate binding sites regions and identifies conserved features in functionally related proteins (proteins that bind the same ligand). The Boundary Cubes method for surface representation was applied to the modified PDB file allowing the creation of graphs for proteins and ligands that could be compared and caused no loss of geometric data. A method is included for describing binding site features of individual ligands conserved in terms of spatial relationships allowed identification of 3D motifs, named fingerprints, which could be searched for in other protein structures. This method combine with a modification of the pocket algorithm allows reduced search areas for graph matching. The methods allow isolation of the binding site from a complexed protein PDB file, identification of conserved features among the binding sites of individual ligand types, and search for these features in sequence data. In terms of spatial conservation create a fingerprint ofthe binding site that can be sought in other proteins of/mown structure, identifYing putative binding sites. The approach offers a novel and generic method for the identification of putative ligand binding sites for proteins for which there is no prior detailed structural characterisation of protein/ ligand interactions. It is unique in being able to convert PDB data into graphs, ready for comparison and thus fitting of ligand to protein with consideration of chemical charge and in the future other chemica! properties.