From Protein Structure to Function with Bioinformatics.pdf

12 Prediction of Protein Function from Theoretical Models 303any defined functionality to the analyzed family. In many cases, however, the proteinstructure can be inferred with the aid of protein fold-recognition methods,alone or in combination with ab initio modelling (Kolinski and Bujnicki 2005) andthen used to pinpoint the potential active site, suggesting a possible function. Thiscan be exemplified by the published analysis (Feder and Bujnicki 2005) of familyof sequences grouped together in the Clusters of Orthologous Groups (COG) database(Tatusov et al. 2003) as COG4636 and annotated as “uncharacterized proteinconserved in Cyanobacteria”. The detailed analysis of sequence conservationwithin COG4636 family combined with secondary structure prediction revealed apattern of α-helices and β-strands associated with conserved carboxylate residues,which has been previously identified in the PD-(D/E)XK superfamily of nucleases(Bujnicki 2003).This similarity suggested that members of COG4636 may belongto the PD-(D/E)XK superfamily (Figs. 12.1a, b). However, the multiple sequenceFig. 12.1 Spatial conservation of the PD-(D/E)XK active site. Only the structurally superimposedcommon cores are shown, terminal regions and insertions have been omitted for clarity of the presentation.The upper panels show bona fide PD-(D/E)XK nucleases – (a) Holliday junctionresolvase Hje (PDB code 1ob8) and (b) restriction endonuclease Ngo-MIV (PDB code 1fiu). Thelower panels present COG4636 structures – (c) a theoretical model (Feder and Bujnicki 2005) and(d) crystal structure of another member of the family (PDB code 1wdj). The side chains of thetypical and variant PD-(D/E)xK active site are coloured orange except the Lys residues which arecoloured blue. This figure was made using PyMOL (http://pymol.sourceforge.net), as were allfigures in this chapter