Observation selection bias in contact prediction and its implications for structural bioinformatics.

Next Generation Sequencing is dramatically increasing the number of known protein sequences, with related experimentally determined protein structures lagging behind. Structural bioinformatics is attempting to close this gap by developing approaches that predict structure-level characteristics for uncharacterized protein sequences, with most of the developed methods relying heavily on evolutionary information collected from homologous sequences. Here we show that there is a substantial observational selection bias in this approach: the predictions are validated on proteins with known structures from the PDB, but exactly for those proteins significantly more homologs are available compared to less studied sequences randomly extracted from Uniprot. Structural bioinformatics methods that were developed this way are thus likely to have over-estimated performances; we demonstrate this for two contact prediction methods, where performances drop up to 60% when taking into account a more realistic amount of evolutionary information. We provide a bias-free dataset for the validation for contact prediction methods called NOUMENON.

PDBe: Protein Data Bank in Europe.

The Protein Data Bank in Europe (PDBe; pdbe.org) is a partner in the Worldwide PDB organization (wwPDB; wwpdb.org) and as such actively involved in managing the single global archive of biomacromolecular structure data, the PDB. In addition, PDBe develops tools, services and resources to make structure-related data more accessible to the biomedical community. Here we describe recently developed, extended or improved services, including an animated structure-presentation widget (PDBportfolio), a widget to graphically display the coverage of any UniProt sequence in the PDB (UniPDB), chemistry- and taxonomy-based PDB-archive browsers (PDBeXplore), and a tool for interactive visualization of NMR structures, corresponding experimental data as well as validation and analysis results (Vivaldi).

PDBe: Protein Data Bank in Europe.

The Protein Data Bank in Europe (PDBe) (http://www.ebi.ac.uk/pdbe/) is actively working with its Worldwide Protein Data Bank partners to enhance the quality and consistency of the international archive of bio-macromolecular structure data, the Protein Data Bank (PDB). PDBe also works closely with its collaborators at the European Bioinformatics Institute and the scientific community around the world to enhance its databases and services by adding curated and actively maintained derived data to the existing structural data in the PDB. We have developed a new database infrastructure based on the remediated PDB archive data and a specially designed database for storing information on interactions between proteins and bound molecules. The group has developed new services that allow users to carry out simple textual queries or more complex 3D structure-based queries. The newly designed 'PDBeView Atlas pages' provide an overview of an individual PDB entry in a user-friendly layout and serve as a starting point to further explore the information available in the PDBe database. PDBe's active involvement with the X-ray crystallography, Nuclear Magnetic Resonance spectroscopy and cryo-Electron Microscopy communities have resulted in improved tools for structure deposition and analysis.

SPINE bioinformatics and data-management aspects of high-throughput structural biology.

SPINE (Structural Proteomics In Europe) was established in 2002 as an integrated research project to develop new methods and technologies for high-throughput structural biology. Development areas were broken down into workpackages and this article gives an overview of ongoing activity in the bioinformatics workpackage. Developments cover target selection, target registration, wet and dry laboratory data management and structure annotation as they pertain to high-throughput studies. Some individual projects and developments are discussed in detail, while those that are covered elsewhere in this issue are treated more briefly. In particular, this overview focuses on the infrastructure of the software that allows the experimentalist to move projects through different areas that are crucial to high-throughput studies, leading to the collation of large data sets which are managed and eventually archived and/or deposited.

E-MSD: improving data deposition and structure quality.

The Macromolecular Structure Database (MSD) (http://www.ebi.ac.uk/msd/) [H. Boutselakis, D. Dimitropoulos, J. Fillon, A. Golovin, K. Henrick, A. Hussain, J. Ionides, M. John, P. A. Keller, E. Krissinel et al. (2003) E-MSD: the European Bioinformatics Institute Macromolecular Structure Database. Nucleic Acids Res., 31, 458-462.] group is one of the three partners in the worldwide Protein DataBank (wwPDB), the consortium entrusted with the collation, maintenance and distribution of the global repository of macromolecular structure data [H. Berman, K. Henrick and H. Nakamura (2003) Announcing the worldwide Protein Data Bank. Nature Struct. Biol., 10, 980.]. Since its inception, the MSD group has worked with partners around the world to improve the quality of PDB data, through a clean up programme that addresses inconsistencies and inaccuracies in the legacy archive. The improvements in data quality in the legacy archive have been achieved largely through the creation of a unified data archive, in the form of a relational database that stores all of the data in the wwPDB. The three partners are working towards improving the tools and methods for the deposition of new data by the community at large. The implementation of the MSD database, together with the parallel development of improved tools and methodologies for data harvesting, validation and archival, has lead to significant improvements in the quality of data that enters the archive. Through this and related projects in the NMR and EM realms the MSD continues to improve the quality of publicly available structural data.

E-MSD: an integrated data resource for bioinformatics.

The Macromolecular Structure Database (MSD) group (http://www.ebi.ac.uk/msd/) continues to enhance the quality and consistency of macromolecular structure data in the Protein Data Bank (PDB) and to work towards the integration of various bioinformatics data resources. We have implemented a simple form-based interface that allows users to query the MSD directly. The MSD 'atlas pages' show all of the information in the MSD for a particular PDB entry. The group has designed new search interfaces aimed at specific areas of interest, such as the environment of ligands and the secondary structures of proteins. We have also implemented a novel search interface that begins to integrate separate MSD search services in a single graphical tool. We have worked closely with collaborators to build a new visualization tool that can present both structure and sequence data in a unified interface, and this data viewer is now used throughout the MSD services for the visualization and presentation of search results. Examples showcasing the functionality and power of these tools are available from tutorial webpages (http://www. ebi.ac.uk/msd-srv/docs/roadshow_tutorial/).

E-MSD: the European Bioinformatics Institute Macromolecular Structure Database.

The E-MSD macromolecular structure relational database (http://www.ebi.ac.uk/msd) is designed to be a single access point for protein and nucleic acid structures and related information. The database is derived from Protein Data Bank (PDB) entries. Relational database technologies are used in a comprehensive cleaning procedure to ensure data uniformity across the whole archive. The search database contains an extensive set of derived properties, goodness-of-fit indicators, and links to other EBI databases including InterPro, GO, and SWISS-PROT, together with links to SCOP, CATH, PFAM and PROSITE. A generic search interface is available, coupled with a fast secondary structure domain search tool.

Conformational model for the consensus V3 loop of the envelope protein gp120 of HIV-1 in a 20% trifluoroethanol/water solution.

Based on experimental NMR data, a model was generated for the conformation of the disulfide-bond-closed cyclic peptide corresponding to the whole V3 loop of the consensus HIV-1 strain in a 20% trifluoroethanol/water solution. The obtained family of structures shows a prominent and well-defined amphipathic alpha helix at the C-terminal end of the peptide from Thr23 to Gln32. A series of turns characterizes the central Gly15-Tyr21 region, while the N-terminal region is poorly defined. Independent experimental data confirms the features of this model, and suggests that this type of conformation can be readily adopted when the V3 loop is in contact with a membrane. The examined V3 loop belongs to a macrophage tropic strain, and using the model, a structural explanation is proposed for the different requirements of V3 loops belonging to macrophage and T-cell line tropic HIV-1 strains.

Design and solution structure of a well-folded stack of two beta-hairpins based on the amino-terminal fragment of human granulin A.

Four amino acid substitutions were introduced into a peptide corresponding to the amino-terminal subdomain (30-31 residues) of human granulin A (HGA) in order to assess the contributions of a hydrophobic framework and other interactions to structure stabilization of the stack of two beta-hairpins. The resulting hybrid peptide, HGA 1-31 (D1V, K3H, S9I, Q20P) with four free cysteines, spontaneously formed a uniquely disulfide-bonded isomer with an expected [1-3, 2-4] disulfide pairing pattern. This peptide was characterized in detail by use of NMR and shown to assume a highly stable structure in solution, in contrast to the amino-terminal 1-30 fragment of human granulin A. The prototype peptide, or HGA 1-30 (C17S, C27S), had lower resistance to chemical reduction and proteolysis, broad NH and H(alpha) proton resonances, lower proton resonance dispersion, and no slowly exchanging amide protons. Two other peptides, HGA 1-30 (C17S, Q20P, C27S) and HGA 1-31 (D1V, K3H, S9I, C17S, C27S), with either Pro20 stabilizing a potential reverse turn or with a hydrophobic cluster consisting of Val1, His3, and Ile9, had sharper and slightly better dispersed NH and H(alpha) proton resonances, but still no slowly exchanging amide protons. The solution structure of HGA 1-31 (D1V, K3H, S9I, Q20P) indicates that it adopts a well-folded conformation of a stack of two beta-hairpins, as found for the amino-terminal subdomain of the prototypic carp granulin-1 with representative beta-hairpin stacks. These results highlight the importance of both hydrophobic and turn-stabilizing interactions for the structural integrity of the hairpin stack scaffold. The conformational stability appears to be maintained by a combination of the well-formed second beta-hairpin and two hydrophobic clusters, one located at the interface between the two beta-hairpins and the other on "top" of the first beta-hairpin. The implications of these findings for the design of conformationally stable hairpin stacks are discussed.

The three-dimensional solution structure of Aesculus hippocastanum antimicrobial protein 1 determined by 1H nuclear magnetic resonance.

Aesculus hippocastanum antimicrobial protein 1 (Ah-AMP1) is a plant defensin isolated from horse chestnuts. The plant defensins have been divided in several subfamilies according to their amino acid sequence homology. Ah-AMP1, belonging to subfamily A2, inhibits growth of a broad range of fungi. So far, a three-dimensional structure has been determined only for members of subfamilies A3 and B2. In order to understand activity and specificity of these plant defensins, the structure of a protein belonging to subfamily A2 is needed. We report the three-dimensional solution structure of Ah-AMP1 as determined from two-dimensional 1H nuclear magnetic resonance data. The structure features all the characteristics of the "cysteine-stabilized alpha beta-motif." A comparison of the structure, the electrostatic potential surface and regions important for interaction with the fungal receptor, is made with Rs-AFP1 (plant defensin of subfamily A3). Thus, residues important for activity and specificity have been assigned.

A 30-residue fragment of the carp granulin-1 protein folds into a stack of two beta-hairpins similar to that found in the native protein.

Upon air oxidation, a peptide corresponding to the 30-residue N-terminal subdomain of carp granulin-1 spontaneously formed the disulfide pairing observed in the native protein. Structural characterization using NMR showed the presence of a defined secondary structure within this peptide. The chemical shifts for most of the alphaCH protons of the peptide and the protein are very similar, and the observed NOE contacts of the peptide strongly resemble those in the protein. A structure calculation of the peptide using NOE distance constraints indicates that the peptide fragment adopts the same conformation as formed within the native protein. The 30-residue N-terminal peptide of carp granulin-1 is the first example of an independently folded stack of two beta-hairpins reinforced by two interhairpin disulfide bonds. Two key areas of the structure show a clustering of hydrophobic residues that may account for its exceptional conformational stability.

Conformation of a Cdc42/Rac interactive binding peptide in complex with Cdc42 and analysis of the binding interface.

Most of the putative effectors for the Rho-family small GTPases Cdc42 and Rac share a common sequence motif referred to as the Cdc42/Rac interactive binding (CRIB) motif. This sequence, with a consensus of I-S-x-P-(x)2-4-F-x-H-x-x-H-V-G [Burbelo, P. D., et al. (1995) J. Biol. Chem. 270, 29071-29074], has been shown to be essential for the functional interactions between these effector proteins and Cdc42. We have characterized the interactions of a 22-residue CRIB peptide derived from human PAK2 [PAK2(71-92)] with Cdc42 using proton and heteronuclear NMR spectroscopy. This CRIB peptide binds to GTP-gammaS-loaded Cdc42 in a saturable manner, with an apparent Kd of 0.6 microM, as determined by fluorescence titration using sNBD-labeled Cdc42. Interaction of the 22-residue peptide PAK2(71-92) with GTP-gammaS-loaded Cdc42 causes resonance perturbations in the 1H-15N HSQC spectrum of Cdc42 that are similar to those observed for a longer (46-amino acid) CRIB-containing protein fragment [Guo, W., et al. (1998) Biochemistry 37, 14030-14037]. Proton NMR studies of PAK2(71-92) demonstrate structuring of PAK2(71-92) in the presence of GTP-gammaS-loaded Cdc42, through the observation of many nonsequential transferred NOEs. Structure calculations based on the observed transferred NOEs show that the central portion of the Cdc42-bound CRIB peptide assumes a loop conformation in which the side chains of consensus residues Phe80, His82, Ile84, His85, and Val86 are brought into proximity. The CRIB motif may therefore represent a minimal interfacial region in the complexes between Cdc42 and its effector proteins.

An NMR-based identification of peptide fragments mimicking the interactions of the cathepsin B propeptide.

Selected fragments of the 62-residue proregion (or residues 1p-62p) of the cysteine protease cathepsin B were synthesized and their interactions with cathepsin B studied by use of proton NMR spectroscopy. Peptide fragments 16p-51p and 26p-51p exhibited differential perturbations of their proton resonances in the presence of cathepsin B. These resonance perturbations were lost for the further truncated 36p-51p fragment, but remained in the 26p-43p and 28p-43p peptide fragments. Residues 23p-26p or TWQ25A in the N-terminal 1p-29p fragment did not show cathepsin B-induced resonance perturbations although the same residues had strongly perturbed proton resonances within the 16p-51p peptide. Both the 1p-29p and 36p-51p fragments lack a common set of hydrophobic residues 30p-35p or F30YNVDI35 from the proregion. The presence of residues F30YNVDI35 appears to confer a conformational preference in peptide fragments 16p-51p, 26p-51p, 28p-43p and 26p-43p, but the same residues induce the aggregation of peptides 16p-36p and 1p-36p. The peptide fragment 26p-43p binds to the active site, as indicated by its inhibition of the catalytic activity of cathepsin B. The cathepsin B prosegment can therefore be reduced into smaller, but functional subunits 28p-43p or 26p-43p that retain specific binding interactions with cathepsin B. These results also suggest that residues F30YNVDI35 may constitute an essential element for the selective inhibition of cathepsin B by the full-length cathepsin B proregion.

Determination of the three-dimensional solution structure of Raphanus sativus antifungal protein 1 by 1H NMR.

Raphanus sativus Antifungal Protein 1 (Rs-AFP1) is a 51 amino acid residue plant defensin isolated from radish (Raphanus sativus L.) seeds. The three-dimensional structure in aqueous solution has been determined from two-dimensional 1H NMR data recorded at 500 MHz using the DIANA/REDAC calculation protocols. Experimental constraints consisted of 787 interproton distances extracted from NOE cross-peaks, 89 torsional constraints from 106 vicinal interproton coupling constants and 32 stereospecific assignments of prochiral protons. Further refinement by simulated annealing resulted in a set of 20 structures having pairwise root-mean-square differences of 1.35(+/- 0.35) A over the backbone heavy atoms and 2.11(+/- 0.46) A over all heavy atoms. The molecule adopts a compact globular fold comprising an alpha-helix from Asn18 till Leu28 and a triple-stranded beta-sheet (beta 1 = Lys2-Arg6, beta 2 = His33-Tyr38 and beta 3 = His43-Pro50). The central strand of this beta-sheet is connected by two disulfide bridges (Cys21-Cys45 and Cys25-Cys47) to the alpha-helix. The connection between beta-strand 2 and 3 is formed by a type VIa beta-turn. Even the loop (Pro7 to Asn17) between beta-strand 1 and the alpha-helix is relatively well defined. The structure of Raphanus sativus Antifungal Protein 1 features all the characteristics of the "cysteine stabilized alpha beta motif". A comparison of the complete structure and of the regions important for interaction with the fungal receptor according to a mutational study, is made with the structure of gamma-thionin, a plant defensin that has no antifungal activity. It is concluded that this interaction is both electrostatic and specific, and some possible scenarios for the mode of action are given.

Conformational features of a synthetic cyclic peptide corresponding to the complete V3 loop of the RF HIV-1 strain in water and water/trifluoroethanol solutions.

The disulfide-bridge-closed cyclic peptide corresponding to the whole V3 loop of the RF HIV-1 strain was examined by proton two-dimensional NMR spectroscopy in water and water/trifluoroethanol solutions. Although most of the peptide is conformationally averaged in water, the NOE data support a beta-turn conformation for the central conservative GPGR region and the presence of nascent helix. Upon addition of trifluoroethanol, helix formation in the C-terminal part becomes apparent. This is confirmed by CD data. NOEs indicative of multiple and transient beta-turns around the Asn6 glycosylation site and NOEs fitting X-ray data on a linear V3 peptide-Fab complex also emerge. The C-terminal helix is shown to have amphipathic character and might thus assist in the infection process.

The complete Consensus V3 loop peptide of the envelope protein gp120 of HIV-1 shows pronounced helical character in solution.

The disulfide bridge closed cyclic peptide corresponding to the whole Consensus V3 loop of the envelope protein gp120 of HIV-1 was examined by proton 2D-NMR spectroscopy in water and in a 20% trifluoroethanol/water solution. In water, NOE data support a beta-turn conformation for the central conservative GPGR region and point towards partial formation of a helix in the C-terminal part. Upon addition of trifluoroethanol, a C-terminal helix is formed. This is evidenced by NOE data, alpha-proton chemical shift changes and changes in the JN alpha vicinal coupling constants. The C-terminal helix is amphipathic and also occurs in other examined strains. It could therefore be an important feature for the functioning of the V3 loop.