Meeting review: the Second meeting on the Critical Assessment of Techniques for Protein Structure Prediction (CASP2), Asilomar, California, December 13-16, 1996

Critical Assessment of Methods for Predicting the 3D Structure of Proteins and Protein Complexes

Advances in a scientific discipline are often measured by small, incremental steps. In this review, we report on two intertwined disciplines in the protein structure prediction field, modeling of single chains and modeling of complexes, that have over decades emulated this pattern, as monitored by the community-wide blind prediction experiments CASP and CAPRI. However, over the past few years, dramatic advances were observed for the accurate prediction of single protein chains, driven by a surge of deep learning methodologies entering the prediction field. We review the mainscientific developments that enabled these recent breakthroughs and feature the important role of blind prediction experiments in building up and nurturing the structure prediction field. We discuss how the new wave of artificial intelligence-based methods is impacting the fields of computational and experimental structural biology and highlight areas in which deep learning methods are likely to lead to future developments, provided that major challenges are overcome.

Assessment of protein assembly prediction in CASP13

We present the assembly category assessment in the 13th edition of the CASP community-wide experiment. For the second time, protein assemblies constitute an independent assessment category. Compared to the last edition we see a clear uptake in participation, more oligomeric targets released, and consistent, albeit modest, improvement of the predictions quality. Looking at the tertiary structure predictions, we observe that ignoring the oligomeric state of the targets hinders modeling success. We also note that some contact prediction groups successfully predicted homomeric interfacial contacts, though it appears that these predictions were not used for assembly modeling. Homology modeling with sizeable human intervention appears to form the basis of the assembly prediction techniques in this round of CASP. Future developments should see more integrated approaches where subunits are modeled in the context of the assemblies they form.

Conformational dynamics of the molecular chaperone Hsp90

The ubiquitous molecular chaperone Hsp90 makes up 1-2% of cytosolic proteins and is required for viability in eukaryotes. Hsp90 affects the folding and activation of a wide variety of substrate proteins including many involved in signaling and regulatory processes. Some of these substrates are implicated in cancer and other diseases, making Hsp90 an attractive drug target. Structural analyses have shown that Hsp90 is a highly dynamic and flexible molecule that can adopt a wide variety of structurally distinct states. One driving force for these rearrangements is the intrinsic ATPase activity of Hsp90, as seen with other chaperones. However, unlike other chaperones, studies have shown that the ATPase cycle of Hsp90 is not conformationally deterministic. That is, rather than dictating the conformational state, ATP binding and hydrolysis only shift the equilibria between a pre-existing set of conformational states. For bacterial, yeast and human Hsp90, there is a conserved three-state (apo-ATP-ADP) conformational cycle; however; the equilibria between states are species specific. In eukaryotes, cytosolic co-chaperones regulate the in vivo dynamic behavior of Hsp90 by shifting conformational equilibria and affecting the kinetics of structural changes and ATP hydrolysis. In this review, we discuss the structural and biochemical studies leading to our current understanding of the conformational dynamics of Hsp90, as well as the roles that nucleotide, co-chaperones, post-translational modification and substrates play. This view of Hsp90's conformational dynamics was enabled by the use of multiple complementary structural methods including, crystallography, small-angle X-ray scattering (SAXS), electron microscopy, Förster resonance energy transfer (FRET) and NMR. Finally, we discuss the effects of Hsp90 inhibitors on conformation and the potential for developing small molecules that inhibit Hsp90 by disrupting the conformational dynamics.

Exploring classification strategies with the CoEPrA 2006 contest

MOTIVATION

In silico methods to classify compounds as potential drugs that bind to a specific target become increasingly important for drug design. To build classification devices training sets of drugs with known activities are needed. For many such classification problems, not only qualitative but also quantitative information of a specific property (e.g. binding affinity) is available. The latter can be used to build a regression scheme to predict this property for new compounds. Predicting a compound property explicitly is generally more difficult than classifying that the property lies below or above a given threshold value. Hence, an indirect classification that is based on regression may lead to poorer results than a direct classification scheme. In fact, initially researchers are only interested to classify compounds as potential drugs. The activities of these compounds are subsequently measured in wet lab.

RESULTS

We propose a novel approach that uses available quantitative information directly for classification rather than first using a regression scheme. It uses a new type of loss function called weighted biased regression. Application of this method to four widely studied datasets of the CoEPrA contest (Comparative Evaluation of Prediction Algorithms, http://coepra.org) shows that it can outperform simple classification methods that do not make use of this additional quantitative information.

AVAILABILITY

A stand alone application is available at the webpage http://agknapp.chemie.fu-berlin.de/agknapp/index.php?menu=software&page=PeptideClassifier that can be used to build a model for a peptide training set to be submitted.

Nanoscale organization of hedgehog is essential for long-range signaling

Hedgehog (Hh) plays crucial roles in tissue-patterning and activates signaling in Patched (Ptc)-expressing cells. Paracrine signaling requires release and transport over many cell diameters away by a process that requires interaction with heparan sulfate proteoglycans (HSPGs). Here, we examine the organization of functional, fluorescently tagged variants in living cells by using optical imaging, FRET microscopy, and mutational studies guided by bioinformatics prediction. We find that cell-surface Hh forms suboptical oligomers, further concentrated in visible clusters colocalized with HSPGs. Mutation of a conserved Lys in a predicted Hh-protomer interaction interface results in an autocrine signaling-competent Hh isoform--incapable of forming dense nanoscale oligomers, interacting with HSPGs, or paracrine signaling. Thus, Hh exhibits a hierarchical organization from the nanoscale to visible clusters with distinct functions.

EGASP: the human ENCODE Genome Annotation Assessment Project

BACKGROUND

We present the results of EGASP, a community experiment to assess the state-of-the-art in genome annotation within the ENCODE regions, which span 1% of the human genome sequence. The experiment had two major goals: the assessment of the accuracy of computational methods to predict protein coding genes; and the overall assessment of the completeness of the current human genome annotations as represented in the ENCODE regions. For the computational prediction assessment, eighteen groups contributed gene predictions. We evaluated these submissions against each other based on a 'reference set' of annotations generated as part of the GENCODE project. These annotations were not available to the prediction groups prior to the submission deadline, so that their predictions were blind and an external advisory committee could perform a fair assessment.

RESULTS

The best methods had at least one gene transcript correctly predicted for close to 70% of the annotated genes. Nevertheless, the multiple transcript accuracy, taking into account alternative splicing, reached only approximately 40% to 50% accuracy. At the coding nucleotide level, the best programs reached an accuracy of 90% in both sensitivity and specificity. Programs relying on mRNA and protein sequences were the most accurate in reproducing the manually curated annotations. Experimental validation shows that only a very small percentage (3.2%) of the selected 221 computationally predicted exons outside of the existing annotation could be verified.

CONCLUSION

This is the first such experiment in human DNA, and we have followed the standards established in a similar experiment, GASP1, in Drosophila melanogaster. We believe the results presented here contribute to the value of ongoing large-scale annotation projects and should guide further experimental methods when being scaled up to the entire human genome sequence.

Automated generation of MCSS-derived pharmacophoric DOCK site points for searching multiconformation databases

All docking methods employ some sort of heuristic to orient the ligand molecules into the binding site of the target structure. An automated method, MCSS2SPTS, for generating chemically labeled site points for docking is presented. MCSS2SPTS employs the program Multiple Copy Simultaneous Search (MCSS) to determine target-based theoretical pharmacophores. More specifically, chemically labeled site points are automatically extracted from selected low-energy functional-group minima and clustered together. These pharmacophoric site points can then be directly matched to the pharmacophoric features of database molecules with the use of either DOCK or PhDOCK to place the small molecules into the binding site. Several examples of the ability of MCSS2SPTS to reproduce the three-dimensional pharmacophoric features of ligands from known ligand-protein complex structures are discussed. In addition, a site-point set calculated for one human immunodeficiency virus 1 (HIV1) protease structure is used with PhDOCK to dock a set of HIV1 protease ligands; the docked poses are compared to the corresponding complex structures of the ligands. Finally, the use of an MCSS2SPTS-derived site-point set for acyl carrier protein synthase is compared to the use of atomic positions from a bound ligand as site points for a large-scale DOCK search. In general, MCSS2SPTS-generated site points focus the search on the more relevant areas and thereby allow for more effective sampling of the target site.

Efficient conformational sampling of local side-chain flexibility

Side-chain flexibility of ligand-binding sites needs to be considered in the rational design of novel inhibitors. We have developed a method to generate conformational ensembles that efficiently sample local side-chain flexibility from a single crystal structure. The rotamer-based approach is tested here for the S1' pocket of human collagenase-1 (MMP-1), which is known to undergo conformational changes in multiple side-chains upon binding of certain inhibitors. First, a raw ensemble consisting of a large number of conformers of the S1' pocket was generated using an exhaustive search of rotamer combinations on a template crystal structure. A combination of principal component analysis and fuzzy clustering was then employed to successfully identify a core ensemble consisting of a low number of representatives from the raw ensemble. The core ensemble contained geometrically diverse conformers of stable nature, as indicated in several cases by a relative energy lower than that of the minimised template crystal structure. Through comparisons with X-ray crystallography and NMR structural data we show that the core ensemble occupied a conformational space similar to that observed under experimental conditions. The synthetic inhibitor RS-104966 is known to induce a conformational change in the side-chains of the S1' pocket of MMP-1 and could not be docked in the template crystal structure. However, the experimental binding mode was reproduced successfully using members of the core ensemble as the docking target, establishing the usefulness of the method in drug design.

Principles of docking: An overview of search algorithms and a guide to scoring functions

The docking field has come of age. The time is ripe to present the principles of docking, reviewing the current state of the field. Two reasons are largely responsible for the maturity of the computational docking area. First, the early optimism that the very presence of the "correct" native conformation within the list of predicted docked conformations signals a near solution to the docking problem, has been replaced by the stark realization of the extreme difficulty of the next scoring/ranking step. Second, in the last couple of years more realistic approaches to handling molecular flexibility in docking schemes have emerged. As in folding, these derive from concepts abstracted from statistical mechanics, namely, populations. Docking and folding are interrelated. From the purely physical standpoint, binding and folding are analogous processes, with similar underlying principles. Computationally, the tools developed for docking will be tremendously useful for folding. For large, multidomain proteins, domain docking is probably the only rational way, mimicking the hierarchical nature of protein folding. The complexity of the problem is huge. Here we divide the computational docking problem into its two separate components. As in folding, solving the docking problem involves efficient search (and matching) algorithms, which cover the relevant conformational space, and selective scoring functions, which are both efficient and effectively discriminate between native and non-native solutions. It is universally recognized that docking of drugs is immensely important. However, protein-protein docking is equally so, relating to recognition, cellular pathways, and macromolecular assemblies. Proteins function when they are bound to other molecules. Consequently, we present the review from both the computational and the biological points of view. Although large, it covers only partially the extensive body of literature, relating to small (drug) and to large protein-protein molecule docking, to rigid and to flexible. Unfortunately, when reviewing these, a major difficulty in assessing the results is the non-uniformity in the formats in which they are presented in the literature. Consequently, we further propose a way to rectify it here.

Reliability of assessment of protein structure prediction methods

The reliability of ranking of protein structure modeling methods is assessed. The assessment is based on the parametric Student's t test and the nonparametric Wilcox signed rank test of statistical significance of the difference between paired samples. The approach is applied to the ranking of the comparative modeling methods tested at the fourth meeting on Critical Assessment of Techniques for Protein Structure Prediction (CASP). It is shown that the 14 CASP4 test sequences may not be sufficient to reliably distinguish between the top eight methods, given the model quality differences and their standard deviations. We suggest that CASP needs to be supplemented by an assessment of protein structure prediction methods that is automated, continuous in time, based on several criteria applied to a large number of models, and with quantitative statistical reliability assigned to each characterization.

On the structure of hisH: protein structure prediction in the context of structural and functional genomics

We predict a structure of the glutamine amidotransferase subunit (hisH) of imidazole glycerol phosphate synthase (IGPS) which catalyzes the fifth step of the histidine biosynthesis in Escherichia coli. The model is constructed using an energy-based threading program augmented by a multiple sequence to structure profile analysis. In developing our model we identified a conserved core region within hisH and a variable domain which is the likely site of interaction with the synthase subunit (hisF) of IGPS. Information available from structural and functional genomics studies was used to improve the structure prediction, to discuss parallels between histidine biosynthesis and other amino acid and nucleotide metabolic pathways, and to better understand the protein-protein interactions between the hisH and hisF domains of IGPS. This work allows us to develop a preliminary model for the structure of the entire IGPS holoenzyme.

Molecular and structural basis of the specificity of a neutralizing acetylcholine receptor-mimicking antibody, using combined mutational and molecular modeling analyses

The antagonist activity of short-chain toxins from snake venoms toward the nicotinic acetylcholine receptor (nAChR) is neutralized upon binding to a toxin-specific monoclonal antibody called Malpha2-3 (1). To establish the molecular basis of this specificity, we predicted from both mutational analyses and docking procedures the structure of the Malpha2-3-toxin complex. From knowledge of the functional paratope and epitope, and using a double-mutation cycle procedure, we gathered evidence that Asp(31) in complementarity determining region 1H is close to, and perhaps interacts with, Arg(33) in the antigen. The use of this pair of proximate residues during the selection procedure yielded three models based on docking calculations. The selected models predicted the proximity of Tyr(49) and/or Tyr(50) in the antibody to Lys(47) in the toxin. This was experimentally confirmed using another round of double-mutation cycles. The two models finally selected were submitted to energy minimization in a CHARMM22 force field, and were characterized by a root mean square deviation of 7.0 +/- 2.9 A. Both models display most features of antibody-antigen structures. Since Malpha2-3 also partially mimics some binding properties of nAChR, these structural features not only explain its fine specificity of recognition, but may also further clarify how toxins bind to nAChR.

Genome annotation assessment in Drosophila melanogaster

Computational methods for automated genome annotation are critical to our community's ability to make full use of the large volume of genomic sequence being generated and released. To explore the accuracy of these automated feature prediction tools in the genomes of higher organisms, we evaluated their performance on a large, well-characterized sequence contig from the Adh region of Drosophila melanogaster. This experiment, known as the Genome Annotation Assessment Project (GASP), was launched in May 1999. Twelve groups, applying state-of-the-art tools, contributed predictions for features including gene structure, protein homologies, promoter sites, and repeat elements. We evaluated these predictions using two standards, one based on previously unreleased high-quality full-length cDNA sequences and a second based on the set of annotations generated as part of an in-depth study of the region by a group of Drosophila experts. Although these standard sets only approximate the unknown distribution of features in this region, we believe that when taken in context the results of an evaluation based on them are meaningful. The results were presented as a tutorial at the conference on Intelligent Systems in Molecular Biology (ISMB-99) in August 1999. Over 95% of the coding nucleotides in the region were correctly identified by the majority of the gene finders, and the correct intron/exon structures were predicted for >40% of the genes. Homology-based annotation techniques recognized and associated functions with almost half of the genes in the region; the remainder were only identified by the ab initio techniques. This experiment also presents the first assessment of promoter prediction techniques for a significant number of genes in a large contiguous region. We discovered that the promoter predictors' high false-positive rates make their predictions difficult to use. Integrating gene finding and cDNA/EST alignments with promoter predictions decreases the number of false-positive classifications but discovers less than one-third of the promoters in the region. We believe that by establishing standards for evaluating genomic annotations and by assessing the performance of existing automated genome annotation tools, this experiment establishes a baseline that contributes to the value of ongoing large-scale annotation projects and should guide further research in genome informatics.

MODBASE, a database of annotated comparative protein structure models

MODBASE is a queryable database of annotated comparative protein structure models. The models are derived by MODPIPE, an automated modeling pipeline relying on the programs PSI-BLAST and MODELLER. The database currently contains 3D models for substantial portions of approximately 17 000 proteins from 10 complete genomes, including those of Caenorhabditis elegans, Saccharomyces cerevisiae and Escherichia coli, as well as all the available sequences from Arabidopsis thaliana and Homo sapiens. The database also includes fold assignments and alignments on which the models were based. In addition, special care is taken to assess the quality of the models. ModBase is accessible through a web interface at http://guitar.rockefeller.edu/modbase/

A systematic study of low-resolution recognition in protein--protein complexes

A comprehensive nonredundant database of 475 cocrystallized protein-protein complexes was used to study low-resolution recognition, which was reported in earlier docking experiments with a small number of proteins. The docking program GRAMM was used to delete the atom-size structural details and systematically dock the resulting molecular images. The results reveal the existence of the low-resolution recognition in 52% of all complexes in the database and in 76% of the 113 complexes with an interface area >4,000 A(2). Limitations of the docking and analysis tools used in this study suggest that the actual number of complexes with the low-resolution recognition is higher. However, the results already prove the existence of the low-resolution recognition on a broad scale.

Side-chain and backbone flexibility in protein core design

We have developed a computational approach for the design and prediction of hydrophobic cores that includes explicit backbone flexibility. The program consists of a two-stage combination of a genetic algorithm and monte carlo sampling using a torsional model of the protein. Backbone structures are evaluated either by a canonical force-field or a constraining potential that emphasizes the preservation of local geometry. The utility of the method for protein design and engineering is explored by designing three novel hydrophobic core variants of the protein 434 cro. We use the new method to evaluate these and previously designed 434 cro variants, as well as a series of phage T4 lysozyme variants. In order to properly evaluate the influence of backbone flexibility, we have also analyzed the effects of varying amounts of side-chain flexibility on the performance of fixed backbone methods. Comparison of results using a fixed versus flexible backbone reveals that, surprisingly, the two methods are almost equivalent in their abilities to predict relative experimental stabilities, but only when full side-chain flexibility is allowed. The prediction of core side-chain structure can vary dramatically between methods. In some, but not all, cases the flexible backbone method is a better predictor of structure. The development of a flexible backbone approach to core design is particularly important for attempts at de novo protein design, where there is no prior knowledge of a precise backbone structure.

Folding funnels, binding funnels, and protein function

Folding funnels have been the focus of considerable attention during the last few years. These have mostly been discussed in the general context of the theory of protein folding. Here we extend the utility of the concept of folding funnels, relating them to biological mechanisms and function. In particular, here we describe the shape of the funnels in light of protein synthesis and folding; flexibility, conformational diversity, and binding mechanisms; and the associated binding funnels, illustrating the multiple routes and the range of complexed conformers. Specifically, the walls of the folding funnels, their crevices, and bumps are related to the complexity of protein folding, and hence to sequential vs. nonsequential folding. Whereas the former is more frequently observed in eukaryotic proteins, where the rate of protein synthesis is slower, the latter is more frequent in prokaryotes, with faster translation rates. The bottoms of the funnels reflect the extent of the flexibility of the proteins. Rugged floors imply a range of conformational isomers, which may be close on the energy landscape. Rather than undergoing an induced fit binding mechanism, the conformational ensembles around the rugged bottoms argue that the conformers, which are most complementary to the ligand, will bind to it with the equilibrium shifting in their favor. Furthermore, depending on the extent of the ruggedness, or of the smoothness with only a few minima, we may infer nonspecific, broad range vs. specific binding. In particular, folding and binding are similar processes, with similar underlying principles. Hence, the shape of the folding funnel of the monomer enables making reasonable guesses regarding the shape of the corresponding binding funnel. Proteins having a broad range of binding, such as proteolytic enzymes or relatively nonspecific endonucleases, may be expected to have not only rugged floors in their folding funnels, but their binding funnels will also behave similarly, with a range of complexed conformations. Hence, knowledge of the shape of the folding funnels is biologically very useful. The converse also holds: If kinetic and thermodynamic data are available, hints regarding the role of the protein and its binding selectivity may be obtained. Thus, the utility of the concept of the funnel carries over to the origin of the protein and to its function.

High-resolution autoreactive epitope mapping and structural modeling of the 65 kDa form of human glutamic acid decarboxylase

The smaller isoform of the GABA-synthesizing enzyme, glutamic acid decarboxylase 65 (GAD65), is unusually susceptible to becoming a target of autoimmunity affecting its major sites of expression, GABA-ergic neurons and pancreatic beta-cells. In contrast, a highly homologous isoform, GAD67, is not an autoantigen. We used homolog-scanning mutagenesis to identify GAD65-specific amino acid residues which form autoreactive B-cell epitopes in this molecule. Detailed mapping of 13 conformational epitopes, recognized by human monoclonal antibodies derived from patients, together with two and three-dimensional structure prediction led to a model of the GAD65 dimer. GAD65 has structural similarities to ornithine decarboxylase in the pyridoxal-5'-phosphate-binding middle domain (residues 201-460) and to dialkylglycine decarboxylase in the C-terminal domain (residues 461-585). Six distinct conformational and one linear epitopes cluster on the hydrophilic face of three amphipathic alpha-helices in exons 14-16 in the C-terminal domain. Two of those epitopes also require amino acids in exon 4 in the N-terminal domain. Two distinct epitopes reside entirely in the N-terminal domain. In the middle domain, four distinct conformational epitopes cluster on a charged patch formed by amino acids from three alpha-helices away from the active site, and a fifth epitope resides at the back of the pyridoxal 5'-phosphate binding site and involves amino acid residues in exons 6 and 11-12. The epitopes localize to multiple hydrophilic patches, several of which also harbor DR*0401-restricted T-cell epitopes, and cover most of the surface of the protein. The results reveal a remarkable spectrum of human autoreactivity to GAD65, targeting almost the entire surface, and suggest that native folded GAD65 is the immunogen for autoreactive B-cells.

Large-scale protein structure modeling of the Saccharomyces cerevisiae genome

The function of a protein generally is determined by its three-dimensional (3D) structure. Thus, it would be useful to know the 3D structure of the thousands of protein sequences that are emerging from the many genome projects. To this end, fold assignment, comparative protein structure modeling, and model evaluation were automated completely. As an illustration, the method was applied to the proteins in the Saccharomyces cerevisiae (baker's yeast) genome. It resulted in all-atom 3D models for substantial segments of 1,071 (17%) of the yeast proteins, only 40 of which have had their 3D structure determined experimentally. Of the 1,071 modeled yeast proteins, 236 were related clearly to a protein of known structure for the first time; 41 of these previously have not been characterized at all.

Protein structure prediction by threading. Why it works and why it does not

We developed a novel Monte Carlo threading algorithm which allows gaps and insertions both in the template structure and threaded sequence. The algorithm is able to find the optimal sequence-structure alignment and sample suboptimal alignments. Using our algorithm we performed sequence-structure alignments for a number of examples for three protein folds (ubiquitin, immunoglobulin and globin) using both "ideal" set of potentials (optimized to provide the best Z-score for a given protein) and more realistic knowledge-based potentials. Two physically different scenarios emerged. If a template structure is similar to the native one (within 2 A RMS), then (i) the optimal threading alignment is correct and robust with respect to deviations of the potential from the "ideal" one; (ii) suboptimal alignments are very similar to the optimal one; (iii) as Monte Carlo temperature decreases a sharp cooperative transition to the optimal alignment is observed. In contrast, if the template structure is only moderately close to the native structure (RMS greater than 3.5 A), then (i) the optimal alignment changes dramatically when an "ideal" potential is substituted by the real one; (ii) the structures of suboptimal alignments are very different from the optimal one, reducing the reliability of the alignment; (iii) the transition to the apparently optimal alignment is non-cooperative. In the intermediate cases when the RMS between the template and the native conformations is in the range between 2 A and 3.5 A, the success of threading alignment may depend on the quality of potentials used. These results are rationalized in terms of a threading free energy landscape. Possible ways to overcome the fundamental limitations of threading are discussed briefly.

Practical evaluation of comparative modelling and threading methods

Six protein pairs, all with known 3D-structures, were used to evaluate different protein structure prediction tools. Firstly, alignments between a target sequence and a template sequence or structure were obtained by sequence alignment with QUANTA or by threading with THREADER, 123D and PHD Topits. Secondly, protein structure models were generated using MODELLER. The two protein structure assessment tools used were the root mean square deviation (RMSD) compared with the experimental target structure and the total 3D profile score. Also the accuracy of the active sites of models built in the absence and presence of ligands was investigated. Our study confirms that threading methods are able to yield more accurate models than comparative modelling in cases of low sequence identity (< 30%). However, a gap of 2 A (RMSD) exists between the theoretically best model and the models obtained by threading methods. For high sequence identities (> 30%) comparative modelling using MODELLER resulted in accurate models. Furthermore, the total 3D profile score was not always able to distinguish correct from incorrect folds when different alignment methods were used. Finally, we found it to be important to include possible ligands in the model-building process in order to prevent unrealistic filling of active site areas.

What is the probability of a chance prediction of a protein structure with an rmsd of 6 A?

BACKGROUND

The root mean square deviation (rmsd) between corresponding atoms of two protein chains is a commonly used measure of similarity between two protein structures. The smaller the rmsd is between two structures, the more similar are these two structures. In protein structure prediction, one needs the rmsd between predicted and experimental structures for which a prediction can be considered to be successful. Success is obvious only when the rmsd is as small as that for closely homologous proteins (< 3 A). To estimate the quality of the prediction in the more general case, one has to compare the native structure not only with the predicted one but also with randomly chosen protein-like folds. One can ask: how many such structures must be considered to find a structure with a given rmsd from the native structure?

RESULTS

We calculated the rmsd values between native structures of 142 proteins and all compact structures obtained in the threading of these protein chains over 364 non-homologous structures. The rmsd distributions have a Gaussian form, with the average rmsd approximately proportional to the radius of gyration.

CONCLUSIONS

We estimated the number of protein-like structures required to obtain a structure within an rmsd of 6 A to be 10(4)-10(5) for chains of 60-80 residues and 10(11)-10(12) structures for chains of 160-200 residues. The probability of obtaining a 6 A rmsd by chance is so remote that when such structures are obtained from a prediction algorithm, it should be considered quite successful.

Assessing protein structures with a non-local atomic interaction energy

We describe a new approach, based on the energy of non-local interactions, to assess protein structures. The method uses a very sensitive and accurate atomic mean force potential (AMFP) to calculate the non-local energy profile (NL-profile) of a proteins structure. Several protein models, built using the comparative modeling technique and containing several errors, were evaluated. These models exhibit a good stereochemistry and have been previously checked with different, widely used, methods that failed to detect the errors. The AMFP-derived energy profiles are able to correlate high scores with point errors and misalignments in the models. The point errors are frequently found in loops or regions of structural differences between the template and the target protein. The misalignments are clearly detected with very high scores. The performance of the method was also tested for the assessment of X-ray solved protein structures. In a data set of 143 well solved and non-redundant protein structures, we find that the average energy Z-scores, obtained from AMFP, increase as the resolution decreases. In the case of structures that have already been described as having an unusual stereochemistry, very high Z-scores are obtained. Moreover, energy calculations for some pairs of obsolete and replacement proteins always show higher Z-scores for the obsolete proteins. Finally, two particular cases show the usefulness of the profiles in the assessment of X-ray solved protein structures. First, the NL-profile of a protein structure refined in the incorrect space group has very high scores in several regions. One region has already been described to be out-of-register with the density map of the structure. The NL-profile of the re-refined structure with the correct space group is vastly improved. In the second case, the method is able to accurately point out disordered residues, even if the atoms of these residues do not violate the sum of the van der Waals radii. ANOLEA, the program used to calculate the NL-profile of a protein structure containing one or more chains is accessible through the World Wide Web at: http://www.fundp.ac.be/pub/ANOLEA.html.

Fold assembly of small proteins using monte carlo simulations driven by restraints derived from multiple sequence alignments

The feasibility of predicting the global fold of small proteins by incorporating predicted secondary and tertiary restraints into ab initio folding simulations has been demonstrated on a test set comprised of 20 non-homologous proteins, of which one was a blind prediction of target 42 in the recent CASP2 contest. These proteins contain from 37 to 100 residues and represent all secondary structural classes and a representative variety of global topologies. Secondary structure restraints are provided by the PHD secondary structure prediction algorithm that incorporates multiple sequence information. Predicted tertiary restraints are derived from multiple sequence alignments via a two-step process. First, seed side-chain contacts are identified from correlated mutation analysis, and then a threading-based algorithm is used to expand the number of these seed contacts. A lattice-based reduced protein model and a folding algorithm designed to incorporate these predicted restraints is described. Depending upon fold complexity, it is possible to assemble native-like topologies whose coordinate root-mean-square deviation from native is between 3.0 A and 6.5 A. The requisite level of accuracy in side-chain contact map prediction can be roughly 25% on average, provided that about 60% of the contact predictions are correct within +/-1 residue and 95% of the predictions are correct within +/-4 residues. Precision in tertiary contact prediction is more critical than absolute accuracy. Furthermore, only a subset of the tertiary contacts, on the order of 25% of the total, is sufficient for successful topology assembly. Overall, this study suggests that the use of restraints derived from multiple sequence alignments combined with a fold assembly algorithm holds considerable promise for the prediction of the global topology of small proteins.

Entamoeba histolytica: computer-assisted modeling of phosphofructokinase for the prediction of broad-spectrum antiparasitic agents

Pyrophosphate-dependent phosphofructokinase (PPi-PFK) is the rate-limiting glycolytic enzyme found in the pathogenic protists Entamoeba histolytica, Giardia lamblia, Toxoplasma gondii, Trichomonas vaginalis, and Naegleria fowleri. The enzyme differs significantly from ATP-dependent phosphofructokinases found in humans and as such represents an important drug target. Current therapy for infections caused by these pathogens is inadequate, especially for children, pregnant women, and the immune compromised. The development of more selective, safer agents in imperative, as parasitic infections are currently a significant health threat worldwide and will likely become increasingly common agents of disease in the future. For the purpose of designing drugs to treat parasitic infections, we have constructed a model of PPi-PFK from E. histolytica based on the three-dimensional structure of the ATP-dependent PFK from Bacillus stearothermophilus. The model was used with the computer program Dock 3.5 (University of California, San Francisco) to predict the binding of pyrophosphate and selected bisphosphonates to the enzyme. The predicted drug-enzyme interactions suggested that two of these compounds would be competitive inhibitors of pyrophosphate. These drugs were tested against E. histolytica and inhibited the growth of amebae in vitro. This class of compounds may have broad-spectrum antiparasitic activity and, in the future, may facilitate the treatment of serious parasitic infections.

Modelling protein docking using shape complementarity, electrostatics and biochemical information

A protein docking study was performed for two classes of biomolecular complexes: six enzyme/inhibitor and four antibody/antigen. Biomolecular complexes for which crystal structures of both the complexed and uncomplexed proteins are available were used for eight of the ten test systems. Our docking experiments consist of a global search of translational and rotational space followed by refinement of the best predictions. Potential complexes are scored on the basis of shape complementarity and favourable electrostatic interactions using Fourier correlation theory. Since proteins undergo conformational changes upon binding, the scoring function must be sufficiently soft to dock unbound structures successfully. Some degree of surface overlap is tolerated to account for side-chain flexibility. Similarly for electrostatics, the interaction of the dispersed point charges of one protein with the Coulombic field of the other is measured rather than precise atomic interactions. We tested our docking protocol using the native rather than the complexed forms of the proteins to address the more scientifically interesting problem of predictive docking. In all but one of our test cases, correctly docked geometries (interface Calpha RMS deviation </=2 A from the experimental structure) are found during a global search of translational and rotational space in a list that was always less than 250 complexes and often less than 30. Varying degrees of biochemical information are still necessary to remove most of the incorrectly docked complexes.

A measure of success in fold recognition

Prediction of protein structure by fold recognition, or threading, was recently put to the test in a 'blind' structure prediction experiment, CASP2. Thirty-two teams from around the world participated, preparing predictions for 22 different 'target' proteins whose structures were soon to be determined. As experimental structures became available, we, as organizers of the threading competition, computed objective measures of fold-recognition specificity and model accuracy, to identify and characterize successful predictions. Here, we present a brief summary of these prediction evaluations, a tally of 'correct' predictions and a discussion of factors associated with correct predictions. We find that threading produced specific recognition and accurate models whenever the structural database contained a template spanning a large fraction of target sequence. Presence of conserved sequence motifs was helpful, but not required, and it would appear that threading can succeed whenever similarity to a known structure is sufficiently extensive.