From Protein Structure to Function with Bioinformatics.pdf

198 E.C. Meng et al.within subtilisins, which contain the catalytic triad in a different fold (Fischer et al.1994). Sequence-order independence is an important feature of many 3D motifmethods; however, in this work, these substructures were relatively large (>50 residues)and were identified from the entire structures rather than being predefined.In other early work in this field, the Thornton group classified catalytic-triadcontainingprotease and lipase structures into four fold groups (Wallace et al. 1996).It was observed that oxygen atoms from the serine and aspartic acid residuesoccupy nearly constant positions relative to the histidine ring across all four groups,whereas the rest of the side chain atoms only superimposed well within each group.An overall 3D motif or template containing just the histidine ring and two oxygenswas generated, along with group-specific templates containing the entire sidechains. To speed comparison, the TESS (TEmplate Search and Superposition) geometrichashing method was developed (Wallace et al. 1997). In this approach, oneresidue in a template provides a frame of reference, the surrounding atoms arebinned into 3D grid cells, and the information is hashed. Structures to be searchedrequire similar pre-processing, where each residue of the same type as the templatereference residue (histidine in the catalytic triad example) is used to define a spatialpattern for hashing. Besides the need for pre-processing and file storage, TESSimposes some limitations on how motifs are defined and matched. The backtrackingconstraint solver JESS (not an acronym) was developed to address these issueswithout sacrificing speed (Barker and Thornton 2003); it performs a depth-firstsearch on efficiently arranged descriptions of structures. The JESS paper alsodescribes obtaining E-values by comparing each 3D motif to a reference set ofstructures and modelling the resulting range of RMSD values as a mixture of normaldistributions (Barker and Thornton 2003).“Fuzzy functional forms” (FFF) consisting of just the alpha-carbons of importantresidues were shown to be useful for screening both experimentally determinedstructures and modeled structures of low to moderate resolution (Fetrow andSkolnick 1998; Di Gennaro et al. 2001). Glutaredoxins/thioredoxins were recognizedwith a motif of two cysteines and a proline, with additional restrictions thatthe proline must be in the cis conformation and the cysteines must be in a CxxCmotif near the N-terminus of a helix. T1 ribonucleases were recognized with a sixresiduemotif. In subsequent work, fuzzy functional forms for identifying broadfamilies were combined with sequence-based active site profiles for finer subclassification(Cammer et al. 2003). The FFF motif for the disulphide oxidoreductaseactive site found in many proteins is shown in Fig. 8.3.The program ASSAM uses subgraph isomorphism to find occurrences of a userdefinedpattern of residues (Artymiuk et al. 1994). Side chain functional groups areeach represented by two or three pseudoatoms, and the distances among thesepoints in a motif are compared to the corresponding distances within a structure.Residues can be labelled by type or chemical classification (for example, hydrophobic).The tradeoff between specificity and distance tolerance was demonstrated forthe catalytic triad, and further examples were presented and discussed. Enhancementsto the original program include the ability to use backbone atoms and to label residuesby secondary structure and degree of solvent exposure (Spriggs et al. 2003).