Protein Structure and Function: Application of Bioinformatics Methods - John Rigden 2014

Spatial motifs
Specific methods
User-defined motifs

Studies on Structure/135.html">Structural motifs can be categorized based on how they address the motif Selection problem. The first group comprises studies focused on evaluating Methods for matching a user-defined motif, leaving The Challenge of selecting an appropriate motif to the method users or future research. The second group consists of works where identifying motifs or creating motif libraries is treated as a major, if not primary, objective of the developed methods.

Methods for identifying user-defined structural motifs are typically demonstrated using a few well-known motifs from literature. The most prominent example is the Ser-His-Asp catalytic triad, first discovered in Serine proteases (Blow et al. 1969; Wright et al. 1969) and later found in other Hydrolases such as esterases and lipases. Serving as an excellent test system due to being thoroughly studied and having A large number of its structures available in the PDB database, this catalytic triad is frequently used to benchmark methods for generating and evaluating structural motifs. The triad occurs across various fold types, thereby encompassing cases of both Divergent and convergent evolution (Fig. 8.2).

When examining serine proteases, geometric hashing using minimal structural information—specifically, only the coordinates of alpha-carbons (rather than residue types or sequence order)—successfully identified not only other serine proteases but also similar substructures in subtilisins, which contain the catalytic triad within a different fold topology (Fischer et al. 1994). Sequence-order independence is a crucial feature of many structural motif methods; however, in the cited study, these substructures were relatively large (>50 residues) and were detected from whole structures rather than being specified in advance.

Class="center">

Fig. 8.2. (For the color version of this figure, see the insert.) Two serine proteases aligned by their catalytic triads exhibit close similarity in Active Site residues despite differences in overall folding, a) Ribbon representation of Trypsin (blue/cyan, PDB ID 1sgt) and proteinase K, a subtilisin homologue (red/pink, PDB ID 2pkc), shows that the two Proteins possess different folds with no corresponding Secondary structure elements except for the catalytic triads (shown in stick representation), which partially overlap. They are believed to share no common evolutionary precursor, b) Side chains of the catalytic triad residues are shown enlarged to highlight the similarity in their orientation (1sgt: Asp102, His57, Ser195; and 2pkc: Asp39, His69, Ser224). The resemblance of the catalytic triad in these non-homologous structures demonstrates the capability of structural motifs to identify analogous Protein Functions in cases where Homology-based methods would fail. Generated using UCSF Chimera visualization software (Pettersen et al. 2004) (http://www.cgl.ucsf.edu/chimera)

In another early contribution to this field, Thornton's group classified protease and lipase structures containing catalytic triads into four groups based on their fold types (Wallace et al. 1996). It was observed that the oxygen atoms of the serine and aspartic acid residues occupy approximately fixed positions relative to the Histidine ring across all four groups, whereas the remaining side-chain atoms align well only within each individual group. A consensus structural motif, or template, comprising solely the histidine ring and two oxygen atoms was constructed, alongside group-specific templates containing the complete side chains. To accelerate the comparison process, the geometric hashing method TESS (Template Search and Superposition) was developed (Wallace et al. 1997).

In this method, a single template residue serves as a reference frame, while surrounding atoms are mapped into spatial grid Cells and the information is hashed. Query structures undergo a similar preprocessing step, in which every residue matching the type of the reference template residue (e.g., the catalytic triad histidine) is used to define a spatial pattern for hashing. Aside from the preprocessing and file storage requirements, TESS imposes certain constraints on motif definition and structural searching. To address these issues without sacrificing computational speed, the JESS algorithm (not an acronym) was developed, implementing constraint satisfaction via backtracking search (Barker and Thornton 2003); the algorithm performs a depth-first search over efficiently ordered structural descriptors. The study also describes deriving expected values by comparing each structural motif against a reference set of structures and modeling the resulting RMSD value ranges as a superposition of normal distributions (Barker and Thornton 2003).

It has been demonstrated that "fuzzy functional forms" (FFFs), consisting of the alpha-carbons of key residues, can be used to screen both experimentally determined and low- or medium-resolution modeled structures (Fetrow and Skolnick 1998; Di Gennaro et al. 2001). Glutaredoxins and thioredoxins were recognized by a motif comprising two cysteines and a Proline, with the additional constraints that proline must be in the cis conformation and cysteines must form a CxxC motif near the N-terminus of the helix. Ribonuclease T1 was identified by a six-residue motif. In subsequent studies, The Use of fuzzy functional forms for recognizing larger families for finer Classification was combined with sequence-based active-site profiling (Cammer et al. 2003). The fuzzy functional form of the disulfide oxidoreductase active-site motif, found in numerous proteins, is illustrated in Fig. 8.3.

The ASSAM program utilizes subgraph isomorphism to search for user-defined residue patterns (Artymiuk et al. 1994). Each side-chain functional group is represented by two or three pseudo-atoms, and the distances between these points in the motif are compared with corresponding distances within the structure. Residues can be labeled either by type or according to chemical classification (e.g., Hydrophobicity). For catalytic triads, the trade-off between Specificity and allowed distance deviation was demonstrated, and additional Examples were presented and discussed. Enhancements to the original program include The ability to utilize backbone atoms and to label residues by secondary structure type and solvent accessibility (Spriggs et al. 2003).

Fig. 8.3. (For the color version of this figure, see the insert.) Fuzzy functional form of the disulfide oxidoreductase active-site motif found in various proteins. The figure displays bacteriophage T4 glutaredoxin, 1aaz, chain A (left), human thioredoxin, 4trx (middle), and disulfide oxidoreductase, 1dsb, chain A (right). The fuzzy functional form is defined by three key residues—two cysteines (side chains shown in red) and a proline (side chains shown in cyan). The active-site structure of these proteins is conserved, although some variations are evident in the rest of the protein.

Using these three key residues, the Characteristic Features of the protein active site were determined (represented by blue ribbons in each protein).

A global sequence alignment of these three proteins, generated using ClustalW, shows the localization of key residues (underlined and highlighted in red and cyan) and the characteristic active-site regions (blue). The alignment illustrates the lack of overall sequence similarity among the three proteins, even though the active-site structure is highly conserved.

In the SPASM program (SPatial Arrangements of Sidechains and Main-chains), each structural motif residue is represented by its alpha-carbon (CA) and/or side-chain center of mass (SC) (Kleywegt 1999). The user specifies which residue types can be mapped to each motif residue, and a exhaustive depth-first search algorithm is employed to identify hits. Pattern alignments for RMSD calculations are preceded by filtering based on CA-CA and SC-SC pseudo-atom distances within each candidate motif. Additionally, sequence-order constraints can be incorporated. Examples include using a three-acid-residue active-site pattern to recognize the glucanase family. Executables and other files for SPASM can be downloaded from the Uppsala Software Factory website (see Table 8.2).

Table 8.2. Web servers for downloading structural motif software

Babbitt's research group utilized SPASM not only with family motifs, where each family is associated with a single function (catalyzed reaction), but also with superfamily motifs linked to a common mechanism shared by one step across various complete reactions (Meng et al. 2004). Motifs derived from single structures enabled the identification of superfamilies with greater sensitivity and specificity than consensus motifs, supporting the hypothesis that coordinate averaging can be detrimental when structures are excessively divergent.

To discover all motifs with RMSD values below a given threshold, the Match Augmentation algorithm performs a priority-driven search starting from the three best-matching structural motif residues and progressively incorporating lower-scoring residues (Chen et al. 2005). Residues were represented as alpha-carbons with labeled residue types and ranked by evolutionary importance derived from sequence alignments (Kristensen et al. 2006), although alternative ranking methods may also be applied. Prioritization reduces the search space, while subsequent performance improvements are achieved through efficient distance comparisons. Match augmentation is carried out using a stack-based depth-first search method. Finally, statistical significance is evaluated using a non-parametric model based on the distribution of RMSD values obtained when a motif is compared against a Background set of protein chains sampled from the PDB. For this purpose, random samples as small as 5% proved sufficient (Chen et al. 2005). It was shown that for motifs consisting of 5–8 points representing isofunctional families (whose members catalyze identical reactions), averaged motif coordinates provide sensitivity comparable to the most sensitive single-structure motifs, alongside a specificity similar to the average specificity of single-structure motifs (Chen et al. 2007b).