From Protein Structure to Function with Bioinformatics.pdf

8 3D Motifs 211its greater overall similarity to the epimerases than to the racemases, the proteinwas found in parallel experimental characterization and in silico docking studies tocatalyze racemization of N-succinyl-Arg/Lys. Furthermore, although the homologymodel was based on an Ala-Glu epimerase structure, docking and rescoring with aphysics-based function and side chain flexibility reproduced the experimentalpreference for N-succinylated basic amino acids. Crystal structures of theprotein in complex with either substrate were obtained and showed remarkableagreement with the docked coordinates. If the side chains were kept rigid, however,docking was much less successful at identifying these substrates (Song et al.2007).In the other study, a structure determined as part of a structural genomicseffort could be identified as a member of the amidohydrolase superfamily basedon its fold and the presence of certain highly conserved active site residues(Hermann et al. 2007). However, the residues were not diagnostic of which (ifany) of the dozens of hydrolysis reactions known for the superfamily might becatalyzed by this particular enzyme. Only metabolites with hydrolysis-susceptiblesubstructures such as amide, ester, and phosphoester groups were chosenfor docking. Because enzymes evolved to catalyze reactions, not just to bindsubstrates, the authors reasoned that using transition-state-like versions of themolecules for docking should provide better predictions. Thus, transition-statelikestructures of each metabolite were generated: amides and esters were elaboratedinto tetrahedral forms, phosphoesters into trigonal bipyramidal forms,and so on. Prior validation studies had shown that using such high-energy formsin docking improved the rankings of known substrates (Hermann et al. 2006).In the prediction work (Hermann et al. 2007), it was noted that many top-scoringcompounds contained an adenine moiety in which the exocyclic nitrogenhad been elaborated into a tetrahedral form, as would be generated duringdeamination. Experimentally, the enzyme was found to catalyze deamination ofthree of four adenine-containing metabolites tested, but not cytosine derivatives,even though the most similar sequences were chlorohydrolases and cytosinedeaminases. A crystal structure of the enzyme in complex with one of thedeamination products showed the same interactions as predicted by docking.Validated functional assignment of this enzyme allowed annotation of severaladditional sequences that also contained the characteristic active site residues.It should be noted that limiting the reaction space to hydrolysis reactions helpedrender the problem tractable, since generating transition-state-like formsgreatly increases the number of molecules to be docked. Scoring includedsteric, electrostatic, and desolvation contributions, but protein flexibility wasnot considered.Three-dimensional motif approaches and docking for functional annotationboth focus on local structure rather than global structure or sequence similarity.While more computationally demanding than 3D motif matching, docking hasthe potential to extrapolate to functions not associated with previouslycharacterized structures; a “new” substrate may be predicted to bind the protein ofinterest.