From Protein Structure to Function with Bioinformatics.pdf

278 J.D. Watson and J.M. Thorntonvariation in the domain that binds the tail of the palmitate, indicating that differentmembers may bind different fatty acids with selectivity for tail length.3. “Twilight zone” proteins. Here neither sequence nor structure strongly suggestsfunction. In the example presented the structure of MJ0226 had a novel fold butshowed weak similarity to nucleotide binding proteins (Hwang et al. 1999).Experimental analysis identified the biochemical function as a novel nucleotidetriphosphatase. In conjunction with observed weak similarities to HAM1 protein(Noskov et al. 1996) the authors suggested a possible role in preventingmutations through removal of non-standard nucleotide triphosphates. This predictionwas later confirmed through a complementation experiment(Stepchenkova et al. 2005).4. New molecular function for known cellular function. Here the overall functionof the protein is known but the biochemical details of its mechanism of actionare revealed by the structure. In the first example quoted, MJ0285 from M. jannaschii,the protein is annotated as being a small heat shock protein inducedunder cellular stress. The structure shows that 24 copies of the protein form ahollow sphere with eight triangular “windows” and six square “windows” (Kimet al. 1998). Based on this information, the question was posed as to whetherpartially denatured proteins are trapped inside the sphere or attached to the outsidesurface. The results of biochemical experiments strongly suggest that thepartially denatured cellular proteins under stress are bound to the outer surfaceof the sphere, thus preventing them from aggregating and inactivating.The second example, MPN625, is a member of the OsmC domain family. Thefamily shows a wide range of sequences, but two highly conserved cysteine residuesare identified by multiple sequence alignment. The crystal structure ofMPN625 revealed that the two conserved cysteine residues lie in the cleft of aputative active site and that this site resembles those in the 2-cysteine peroxiredoxinfamily whose function is to inactivate reactive oxygen species (Schroderet al. 2000). Therefore the comparison of these active sites combined withknown cellular function provided hints to the molecular function of this proteinfamily and an explanation for the differences in substrate specificity.5. Proteins where their function remains unknown. Here the two examples quoted,Aq1575 from Aquifex aeolicus and MPN314 from Mycoplasma pneumoniae,are both hypothetical proteins which are members of Pfam domains of unknownfunction. In both cases there is evidence from conserved residues to suggest aputative active site, but searches of all the motif and functional databases failedto provide any clues as to the molecular function of these proteins.Perhaps the largest analysis performed to date is that of Watson et al. (2007) whichexamined the effectiveness of the ProFunc server for structure-based function predictionusing structures solved by the Midwest Center for Structural Genomics(MCSG). In this study, all 319 structures solved by the MCSG during the first stageof the NIH/NIGMS Protein Structure Initiative (PSI-1) were classified into thosewith known function, those with putative functions assigned and those of unknownfunction. Only those proteins of known function were examined, as the aim was to