Three-dimensional searching for recurrent structural motifs in data bases of protein structures

Super: a web server to rapidly screen superposable oligopeptide fragments from the protein data bank

Searching for well-fitting 3D oligopeptide fragments within a large collection of protein structures is an important task central to many analyses involving protein structures. This article reports a new web server, Super, dedicated to the task of rapidly screening the protein data bank (PDB) to identify all fragments that superpose with a query under a prespecified threshold of root-mean-square deviation (RMSD). Super relies on efficiently computing a mathematical bound on the commonly used structural similarity measure, RMSD of superposition. This allows the server to filter out a large proportion of fragments that are unrelated to the query; >99% of the total number of fragments in some cases. For a typical query, Super scans the current PDB containing over 80,500 structures (with ∼40 million potential oligopeptide fragments to match) in under a minute. Super web server is freely accessible from: http://lcb.infotech.monash.edu.au/super.

Characterizing the regularity of tetrahedral packing motifs in protein tertiary structure

MOTIVATION

While protein secondary structure is well understood, representing the repetitive nature of tertiary packing in proteins remains difficult. We have developed a construct called the relative packing group (RPG) that applies the clique concept from graph theory as a natural basis for defining the packing motifs in proteins. An RPG is defined as a clique of residues, where every member contacts all others as determined by the Delaunay tessellation. Geometrically similar RPGs define a regular element of tertiary structure or tertiary motif (TerMo). This intuitive construct provides a simple approach to characterize general repetitive elements of tertiary structure.

RESULTS

A dataset of over 4 million tetrahedral RPGs was clustered using different criteria to characterize the various aspects of regular tertiary structure in TerMos. Grouping this data within the SCOP classification levels of Family, Superfamily, Fold, Class and PDB showed that similar packing is shared across different folds. Classification of RPGs based on residue sequence locality reveals topological preferences according to protein sizes and secondary structure. We find that larger proteins favor RPGs with three local residues packed against a non-local residue. Classifying by secondary structure, helices prefer mostly local residues, sheets favor at least two local residues, while turns and coil populate with more local residues. To depict these TerMos, we have developed 2 complementary and intuitive representations: (i) Dirichlet process mixture density estimation of the torsion angle distributions and (ii) kernel density estimation of the Cartesian coordinate distribution. The TerMo library and representations software are available upon request.

MUSTANG: a multiple structural alignment algorithm

Multiple structural alignment is a fundamental problem in structural genomics. In this article, we define a reliable and robust algorithm, MUSTANG (MUltiple STructural AligNment AlGorithm), for the alignment of multiple protein structures. Given a set of protein structures, the program constructs a multiple alignment using the spatial information of the C(alpha) atoms in the set. Broadly based on the progressive pairwise heuristic, this algorithm gains accuracy through novel and effective refinement phases. MUSTANG reports the multiple sequence alignment and the corresponding superposition of structures. Alignments generated by MUSTANG are compared with several handcurated alignments in the literature as well as with the benchmark alignments of 1033 alignment families from the HOMSTRAD database. The performance of MUSTANG was compared with DALI at a pairwise level, and with other multiple structural alignment tools such as POSA, CE-MC, MALECON, and MultiProt. MUSTANG performs comparably to popular pairwise and multiple structural alignment tools for closely related proteins, and performs more reliably than other multiple structural alignment methods on hard data sets containing distantly related proteins or proteins that show conformational changes.

Conformations of the third hypervariable region in the VH domain of immunoglobulins

Antigen-combining sites of antibodies are constructed from six loops from VL and VH domains. The third hypervariable region of the heavy chain is far more variable than the others in length, sequence and structure, and was not included in the canonical-structure description of the conformational repertoire of the three hypervariable regions of V kappa chains and the first two of VH chains. Here we present an analysis of the conformations of the third hypervariable region of VH domains (the H3 regions) in antibodies of known structure. We define the H3 region as comprising the residues between 92Cys and 104Gly. We divide it into a torso comprising residues proximal to the framework, four residues from the N terminus and six residues from the C terminus, and a head. There are two major classes of H3 structures that have more than ten residues between 92Cys and 104Gly: (1) the conformation of the torso has a beta-bulge at residue 101, and (2) the torso does not contain a bulge, but continues the regular hydrogen-bonding pattern of the beta-sheet hairpin. The choice of bulged versus non-bulged torso conformation is dictated primarily by the sequence, through the formation of a salt bridge between the side-chains of an Arg or Lys at position 94 and an Asp at position 101. Thus the torso region appears to have a limited repertoire of conformations, as in the canonical structure model of other antigen-binding loops. The heads or apices of the loops have a very wide variety of conformations. In shorter H3 regions, and in those containing the non-bulged torso conformation, the heads follow the rules relating sequence to structure in short hairpins. We surveyed the heads of longer H3 regions, finding that those with bulged torsos present many very different conformations of the head. We recognize that H3, unlike the other five antigen-binding loops, has a conformation that depends strongly on the environment, and we have analysed the interactions of H3 with residues elsewhere in the VH domain, in the VL domain, and with ligands, and their effects on the conformation of H3. We tested these results by attempts to predict the conformations of H3 regions in antibody structures solved after the results were derived. The general conclusion of this work is that the conformation of H3 shows some regularities, from which rules relating sequence to conformation can be stated, but to a less complete degree than for the other five antigen-binding loops. Accurate prediction of the torso conformation is possible in most cases; predictions of the conformation of the head is possible in some cases. However, our understanding of the sequence-structure relationships has reduced the uncertainty to no more than a few residues at the apex of the H3 region.

Integrated access to genomic and other bioinformation: an essential ingredient of the drug discovery process

Due to the high rate of data production and the need of researchers to have rapid access to new data, public databases have become the major medium through which genome mapping and sequencing data as well as macromolecular structural data are published. There are now more than 250 databases of biomolecular, structural, genetic, or phenotypic data, many of which are doubling in size annually. These databases, many of which were created and are maintained by experimentalists for their own research use, provide valuable collections of organized, validated data. However, the very number and diversity of databases now make efficient data resource discovery as important as effective data resource use. Existing autonomous biological databases contain related data which are more valuable when interconnected than when isolated. Political and scientific realities dictate that these databases will be built by different teams, in different locations, for different purposes, and using different data models and supporting DBMSs. As a consequence, connecting the related data they contain is not straightforward. Experience with existing biological databases indicates that it is possible to form useful queries across these databases, but that doing so usually requires expertise in the semantic structure of each source database. Advancing to the next level of integration among biological information resources poses significant technical and sociological challenges.

Antibody structure, prediction and redesign

So far the difficulty to predict the structure of the third hypervariable loop of the heavy chain of antibodies has represented the main limitation in modelling the complete antigen binding site. We carefully analysed all available structures of immunoglobulins searching for rules relating the loop conformation to its amino acid sequence. Here, we analyse the conformation of this loop and show that we are able to predict the conformation of the ten residues proximal to the framework. The conformation of the remaining residues of loops longer than 10 residues can also be predicted in many cases. This, combined with the previously defined canonical structures for the other five hypervariable loops, is an important step toward the prediction of the complete immunoglobulin antigen-binding site. We exemplify our prediction protocol using three known immunoglobulin structures as test cases.

Extraction of well-fitting substructures: root-mean-square deviation and the difference distance matrix

The extraction of well-fitting substructures of two or more sets of proteins has applications to analysis of mechanisms of conformational change in proteins, including pathways of evolution, to classification of protein folding patterns, and to evaluation of protein structure predictions. Many methods are known for extracting some substantial common substructure with low root-mean-square deviation (r.m.s.d.). A harder problem is addressed here: finding all common substructures with r.m.s.d. less than a prespecified threshold. Our approach is to consider the minimum value of the maximum distance between corresponding points, corresponding to superposition in the Chebyshev norm. Using the properties of Chebyshev superposition, we derive relationships between the r.m.s.d. and the maximum element of the difference matrix, two common measures of structural similarity. The results provide a basis for developing algorithms and software to identify all well-fitting subsets.

Bioinformatics--principles and potential of a new multidisciplinary tool

The materials of bioinformatics are biological data, and its methods are derived from a wide variety of computational techniques. Recent years have seen an explosive growth in biological data, and the development of novel computational methods. These methods have become essential to research progress in structural biology, genomics, structure-based drug design and molecular evolution. The development and maintenance of a robust infrastructure of biological data is of equal importance if biotechnology is to take maximum advantage of research advances in a wide variety of fields. While bioinformatics has already made important contributions, it faces significant challenges as it matures.