From Protein Structure to Function with Bioinformatics.pdf

202 E.C. Meng et al.Oldfield analyzed a representative set of structures by excluding small nonpolarresidues, treating residues as single points, collating triplets of residues intogroups of like types, and for each group, sorting interresidue distances into bins0.5 Å wide (Oldfield 2002). Highly populated bins in the resulting 3D histogram(a triplet has three interresidue distances) represent frequent patterns of that tripletof types. Frequent patterns were coalesced as possible to include more than threeresidues. The process identified several known 3D motifs such as binding sites andthe catalytic triad. Programs for comparing structures to the motifs were alsodescribed (Oldfield 2002).Another study involved all-by-all pairwise comparisons of likely functional sitescomprised of residues lining cavities and which were either near a ligand or conservedin sequence alignments (Ausiello et al. 2007). While directed at sitesexpected to be important for function, this mining was still undirected in that thestructures were not grouped by any structural or functional criteria. To identifycases of convergent evolution, the authors focused on matches in which the pairedresidues were in different orders in the respective sequences. Known examples ofsuch permutations were identified, as well as a novel case. Matching sets of residueswere found with the Query3D program (Ausiello et al. 2005a), which performsan exhaustive depth-first search using two points per residue, alpha-carbonand side chain centroid. Query3D identifies matches of up to ten residue pairs,where paired residues are of similar types and the entire match superimposes withan RMSD value below a cutoff.8.3.2.3 Individual StructuresSeveral databases of 3D motifs have been generated using only information fromeach source structure individually. For example, binding site motifs can be collectedby taking residues within a cutoff distance of ligands, nucleic acids, or evenother protein chains. Often, these studies focus on the search tools rather than onthe databases, and some also present results of other types of searches, includingsearches with motifs taken from published literature.The PINTS (Patterns in Non-homologous Tertiary Structures) server (Stark andRussell 2003) (Table 8.1) compares a query structure to a database of 3D motifs,either binding sites defined as residues within 3 Å of a ligand, or SITE motifs basedon PDB SITE annotations. Alternatively, a user-defined motif can be compared toa database of proteins (such as representatives at different SCOP levels), or twospecified structures can be compared to each other. Similar to Russell’s earlierundirected mining work (Russell 1998), PINTS performs a depth-first search anddisregards nonpolar residues. The server uses side chain atoms and allows certainsimilar residue types to match. Statistical significance is estimated with a methoddeveloped by the authors (Stark et al. 2003), described above for the CSA. Resultsof weekly comparisons of newly deposited PDB structures to the motif databasesare also available at the PINTS web site (Stark et al. 2004) (Table 8.1).