A decade of CASP: progress, bottlenecks and prognosis in protein structure prediction

For the past ten years, CASP (Critical Assessment of Structure Prediction) has monitored the state of the art in modeling protein structure from sequence. During this period, there has been substantial progress in both comparative modeling of structure (using information from an evolutionarily related structural template) and template-free modeling. The quality of comparative models depends on the closeness of the evolutionary relationship on which they are based. Template-free modeling, although still very approximate, now produces topologically near correct models for some small proteins. Current major challenges are refining comparative models so that they match experimental accuracy, obtaining accurate sequence alignments for models based on remote evolutionary relationships, and extending template-free modeling methods so that they produce more accurate models, handle parts of comparative models not available from a template and deal with larger structures.     

A comparative study of available software for high-accuracy homology modeling: from sequence alignments to structural models

An open question in protein homology modeling is, how well do current modeling packages satisfy the dual criteria of quality of results and practical ease of use? To address this question objectively, we examined homology‐built models of a variety of therapeutically relevant proteins. The sequence identities across these proteins range from 19% to 76%. A novel metric, the difference alignment index (DAI), is developed to aid in quantifying the quality of local sequence alignments. The DAI is also used to construct the relative sequence alignment (RSA), a new representation of global sequence alignment that facilitates comparison of sequence alignments from different methods. Comparisons of the sequence alignments in terms of the RSA and alignment methodologies are made to better understand the advantages and caveats of each method. All sequence alignments and corresponding 3D models are compared to their respective structure‐based alignments and crystal structures. A variety of protein modeling software was used. We find that at sequence identities >40%, all packages give similar (and satisfactory) results; at lower sequence identities (<25%), the sequence alignments generated by Profit and Prime, which incorporate structural information in their sequence alignment, stand out from the rest. Moreover, the model generated by Prime in this low sequence identity region is noted to be superior to the rest. Additionally, we note that DSModeler and MOE, which generate reasonable models for sequence identities >25%, are significantly more functional and easier to use when compared with the other structure‐building software.     

Strategies for high-throughput comparative modeling: applications to leverage analysis in structural genomics and protein family organization

The technological breakthroughs in structural genomics were designed to facilitate the solution of a sufficient number of structures, so that as many protein sequences as possible can be structurally characterized with the aid of comparative modeling. The leverage of a solved structure is the number and quality of the models that can be produced using the structure as a template for modeling and may be viewed as the "currency" with which the success of a structural genomics endeavor can be measured. Moreover, the models obtained in this way should be valuable to all biologists. To this end, at the Northeast Structural Genomics Consortium (NESG), a modular computational pipeline for automated high-throughput leverage analysis was devised and used to assess the leverage of the 186 unique NESG structures solved during the first phase of the Protein Structure Initiative (January 2000 to July 2005). Here, the results of this analysis are presented. The number of sequences in the nonredundant protein sequence database covered by quality models produced by the pipeline is approximately 39,000, so that the average leverage is approximately 210 models per structure. Interestingly, only 7900 of these models fulfill the stringent modeling criterion of being at least 30% sequence-identical to the corresponding NESG structures. This study shows how high-throughput modeling increases the efficiency of structure determination efforts by providing enhanced coverage of protein structure space. In addition, the approach is useful in refining the boundaries of structural domains within larger protein sequences, subclassifying sequence diverse protein families, and defining structure-based strategies specific to a particular family.     

Predicting protein three-dimensional structure

The current state of the art in modeling protein structure has been assessed, based on the results of the CASP (Critical Assessment of protein Structure Prediction) experiments. In comparative modeling, improvements have been made in sequence alignment, sidechain orientation and loop building. Refinement of the models remains a serious challenge. Improved sequence profile methods have had a large impact in fold recognition. Although there has been some progress in alignment quality, this factor still limits model usefulness. In ab initio structure prediction, there has been notable progress in building approximately correct structures of 40-60 residue-long protein fragments. There is still a long way to go before the general ab initio prediction problem is solved. Overall, the field is maturing into a practical technology, able to deliver useful models for a large number of sequences.     

Modeling repetitive,non‐globular proteins

While ab initio modeling of protein structures is not routine, certain types of proteins are more straightforward to model than others. Proteins with short repetitive sequences typically exhibit repetitive structures. These repetitive sequences can be more amenable to modeling if some information is known about the predominant secondary structure or other key features of the protein sequence. We have successfully built models of a number of repetitive structures with novel folds using knowledge of the consensus sequence within the sequence repeat and an understanding of the likely secondary structures that these may adopt. Our methods for achieving this success are reviewed here.     

Protein structure prediction

The prediction of protein structure, based primarily on sequence and structure homology, has become an increasingly important activity. Homology models have become more accurate and their range of applicability has increased. Progress has come, in part, from the flood of sequence and structure information that has appeared over the past few years, and also from improvements in analysis tools. These include profile methods for sequence searches, the use of three-dimensional structure information in sequence alignment and new homology modeling tools, specifically in the prediction of loop and side-chain conformations. There have also been important advances in understanding the physical chemical basis of protein stability and the corresponding use of physical chemical potential functions to identify correctly folded from incorrectly folded protein conformations.     

How inaccuracies in protein structure models affect estimates of protein-ligand interactions: computational analysis of HIV-I protease inhibitor binding

The influence of possible inaccuracies that can arise during homology modeling of protein structures used for ligand binding studies were investigated with the molecular mechanics generalized Born surface area (MM-GBSA) method. For this, a family of well-characterized HIV-I protease-inhibitor complexes was used. Validation of MM-GBSA led to a correlation coefficient ranging from 0.72 to 0.93 between calculated and experimental binding free energies DeltaG. All calculated DeltaG values were based on molecular dynamics simulations with explicit solvent. Errors introduced into the protein structure through misplacement of side-chains during rotamer modeling led to a correlation coefficient between DeltaG(calc) and DeltaG(exp) of 0.75 compared with 0.90 for the correctly placed side chains. This is in contrast to homology models for members of the retroviral protease family with template structures ranging in sequence identity between 32% and 51%. For these protein models, the correlation coefficients vary between 0.84 and 0.87, which is considerably closer to the original protein (0.90). It is concluded that HIV-I low sequence identity with the template structure still allows creating sufficiently reliable homology models to be used for ligand-binding studies, although placement of the rotamers is a critical step during the modeling.     

Assessment of the ability to model proteins with leucine-rich repeats in light of the latest structural information

The three-dimensional structures of leucine-rich repeat (LRR)-containing proteins from five different families were previously predicted based on the crystal structure of the ribonuclease inhibitor, using an approach that combined homology-based modeling, structure-based sequence alignment of LRRs, and several rational assumptions. The structural models have been produced based on very limited sequence similarity, which, in general, cannot yield trustworthy predictions. Recently, the protein structures from three of these five families have been determined. In this report we estimate the quality of the modeling approach by comparing the models with the experimentally determined structures. The comparison suggests that the general architecture, curvature, "interior/exterior" orientations of side chains, and backbone conformation of the LRR structures can be predicted correctly. On the other hand, the analysis revealed that, in some cases, it is difficult to predict correctly the twist of the overall super-helical structure. Taking into consideration the conclusions from these comparisons, we identified a new family of bacterial LRR proteins and present its structural model. The reliability of the LRR protein modeling suggests that it would be informative to apply similar modeling approaches to other classes of solenoid proteins.     

Refinement of unreliable local regions in template-based protein models

Contemporary template-based modeling techniques allow applications of modeling methods to vast biological problems. However, they tend to fail to provide accurate structures for less-conserved local regions in sequence even when the overall structure can be modeled reliably. We call these regions unreliable local regions (ULRs). Accurate modeling of ULRs is of enormous value because they are frequently involved in functional specificity. In this article, we introduce a new method for modeling ULRs in template-based models by employing a sophisticated loop modeling technique. Combined with our previous study on protein termini, the method is applicable to refinement of both loop and terminus ULRs. A large-scale test carried out in a blind fashion in CASP9 (the 9th Critical Assessment of techniques for protein structure prediction) shows that ULR structures are improved over initial template-based models by refinement in more than 70% of the successfully detected ULRs. It is also notable that successful modeling of several long ULRs over 12 residues is achieved. Overall, the current results show that a careful application of loop and terminus modeling can be a promising tool for model refinement in template-based modeling.     

Revisiting the homology modeling of G-protein coupled receptors: β1-adrenoceptor as an example

G-protein coupled receptors (GPCRs) are recognized to constitute the largest family of membrane proteins. Due to the disproportion in the quantity of crystal structures and their amino acid sequences, homology modeling contributes a reasonable and feasible approach to GPCR theoretical coordinates. With the brand new crystal structures resolved recently, herein we deliberated how to designate them as templates to carry out homology modeling in four aspects: (1) various sequence alignment methods; (2) protein weight matrix; (3) different sets of multiple templates; (4) active and inactive state of templates. The accuracy of models was evaluated by comparing the similarity of stereo conformation and molecular docking results between models and the experimental structure of Meleagris gallopavo β(1)-adrenergic receptor (Mg_Adrb1) that we desired to develop as an example. Our results proposed that: (1) Cobalt and MAFFT, two algorithms of sequence alignment, were suitable for single- and multiple-template modeling, respectively; (2) Blosum30 is applicable to align sequences in the case of low sequence identity; (3) multiple-template modeling is not always better than single-template one; (4) the state of template is an influential factor in simulating the GPCR structures as well.     

Comparative modelling of tetramanganese cluster of Chaetosphaeridium globosum

The DI protein of photosystem II (PS II) complex of a microalga Chaetosphaeridium globosum has been theoretically modelled from its sequence using comparative modeling with known backbone structure of DI protein from bacterium Thermosynechococcus vulcanus as template. The model is built with missing loops and all side chains, which are not resolved in the structure of the template. The structure of the tetramanganese cluster (TMC) and the ligand forming side chains have been subjected to modeling studies in order to gather more information useful to understanding of the water splitting reactions. Earlier models of TMC have been scrutinized and an insight into the manganese coordination sphere has been provided.     

基于结构比较的蛋白质模建系统及其评估

对蛋白质三维结构的同源模建的所有关键模块作了详细研究后,现已整合成蛋白质结构模建的软件系统ＰＭＯＤＥＬＬＩＮＧ。它分主链模建和侧链安装两部分,共七个模块组成。经１１族同源蛋白的模建实验的回顾性检表明,与目前流行的商品软件相比,具有操作自动化,模建精度高和运行时间少的优点。文中还对蝮蛇蛇毒磷酯酶,细胞因子受体ＣＤ４０的配体ｇｐ３９,小鼠巨细胞蛋白酶和人白介素１３等至今未知结构的蛋白作了预测性模建。     

Crystal Structure and Computational Modeling of the Fab Fragment from a Protective Anti-Ricin Monoclonal Antibody

Background

Many antibody crystal structures have been solved. Structural modeling programs have been developed that utilize this information to predict 3-D structures of an antibody based upon its sequence. Because of the problem of self-reference, the accuracy and utility of these predictions can only be tested when a new structure has not yet been deposited in the Protein Data Bank.

Methods

We have solved the crystal structure of the Fab fragment of RAC18, a protective anti-ricin mAb, to 1.9 Å resolution. We have also modeled the Fv structure of RAC18 using publicly available Ab modeling tools Prediction of Immunoglobulin Structures (PIGS), RosettaAntibody, and Web Antibody Modeling (WAM). The model structures underwent energy minimization. We compared results to the crystal structure on the basis of root-mean-square deviation (RMSD), template modeling score (TM-score), Z-score, and MolProbity analysis.

Findings

The crystal structure showed a pocket formed mainly by AA residues in each of the heavy chain complementarity determining regions (CDRs). There were differences between the crystal structure and structures predicted by the modeling tools, particularly in the CDRs. There were also differences among the predicted models, although the differences were small and within experimental error. No one modeling program was clearly superior to the others. In some cases, choosing structures based only on sequence homology to the crystallized Ab yielded RMSDs comparable to the models.

Conclusions

Molecular modeling programs accurately predict the structure of most regions of antibody variable domains of RAC18. The hypervariable CDRs proved most difficult to model, particularly H chain CDR3. Because CDR3 is most often involved in contact with antigen, this defect must be considered when using models to identify potential contacts between antibody and antigen. Because this study represents only a single case, the results cannot be generalized. Rather they highlight the utility and limitations of modeling programs.     

A multiple-template approach to protein threading

Most threading methods predict the structure of a protein using only a single template. Due to the increasing number of solved structures, a protein without solved structure is very likely to have more than one similar template structures. Therefore, a natural question to ask is if we can improve modeling accuracy using multiple templates. This article describes a new multiple-template threading method to answer this question. At the heart of this multiple-template threading method is a novel probabilistic-consistency algorithm that can accurately align a single protein sequence simultaneously to multiple templates. Experimental results indicate that our multiple-template method can improve pairwise sequence-template alignment accuracy and generate models with better quality than single-template models even if they are built from the best single templates (P-value <10(-6)) while many popular multiple sequence/structure alignment tools fail to do so. The underlying reason is that our probabilistic-consistency algorithm can generate accurate multiple sequence/template alignments. In another word, without an accurate multiple sequence/template alignment, the modeling accuracy cannot be improved by simply using multiple templates to increase alignment coverage. Blindly tested on the CASP9 targets with more than one good template structures, our method outperforms all other CASP9 servers except two (Zhang-Server and QUARK of the same group). Our probabilistic-consistency algorithm can possibly be extended to align multiple protein/RNA sequences and structures.     

Structural and functional conservation profiles of novel cathepsin L-like proteins identified in the Drosophila melanogaster genome

Cathepsin L is a cysteine protease which degrades connective tissue proteins including collagen, elastin, and fibronectin. In this study, five well-characterized cathepsin L proteins from different arthropods were used as query sequences for the Drosophila genome database. The search yielded 10 cathepsin L-like sequences, of which eight putatively represent novel cathepsin L-like proteins. To understand the phylogenetic relationship among these cathepsin L-like proteins, a phylogenetic tree was constructed based on their sequences. In addition, models of the tertiary structures of cathepsin L were constructed using homology modeling methods and subjected to molecular dynamics simulations to obtain reasonable structure to understand its dynamical behavior. Our findings demonstrate that all of the potential Drosophila cathepsin L-like proteins contain at least one cathepsin propeptide inhibitor domain. Multiple sequence alignment and homology models clearly highlight the conservation of active site residues, disulfide bonds, and amino acid residues critical for inhibitor binding. Furthermore, comparative modeling indicates that the sequence/structure/function profiles and active site architectures are conserved.     

Bridging the gap in RNA structure prediction

The field of RNA structure prediction has experienced significant advances in the past several years, thanks to the availability of new experimental data and improved computational methodologies. These methods determine RNA secondary structures and pseudoknots from sequence alignments, thermodynamics-based dynamic programming algorithms, genetic algorithms and combined approaches. Computational RNA three-dimensional modeling uses this information in conjunction with manual manipulation, constraint satisfaction methods, molecular mechanics and molecular dynamics. The ultimate goal of automatically producing RNA three-dimensional models from given secondary and tertiary structure data, however, is still not fully realized. Recent developments in the computational prediction of RNA structure have helped bridge the gap between RNA secondary structure prediction, including pseudoknots, and three-dimensional modeling of RNA.     

Structural conservation in single-domain proteins: implications for homology modeling

Large-scale sequencing projects are widening the gap between the known protein universe and the fraction for which structural information has been experimentally obtained. Through the application of homology (comparative) modeling and more general structure prediction techniques, this gap can, however, be narrowed, providing indirect structural information for a considerable number of proteins. Moreover, the estimated number of existing protein folds seems to be limited and many of these yet unknown folds should be discovered by dedicated large-scale structural genomics projects. Within this perspective, homology (comparative) modeling will gain in importance, as will the use of models derived by this technique. Here we discuss how well a sequence alignment, the most common starting point for generating a model, reflects the structural conservation between homologous proteins and we show that sequence information is able to direct construction of acceptable models as far as the structural core is concerned. We also show here that the regions surrounding insertions and deletions are much less conserved than the core and discuss the implications of this observation for loop modeling.     

Assessing model accuracy using the homology modeling automatically software

Homology modeling is a powerful technique that greatly increases the value of experimental structure determination by using the structural information of one protein to predict the structures of homologous proteins. We have previously described a method of homology modeling by satisfaction of spatial restraints (Li et al., Protein Sci 1997;6:956-970). The Homology Modeling Automatically (HOMA) web site, , is a new tool, using this method to predict 3D structure of a target protein based on the sequence alignment of the target protein to a template protein and the structure coordinates of the template. The user is presented with the resulting models, together with an extensive structure validation report providing critical assessments of the quality of the resulting homology models. The homology modeling method employed by HOMA was assessed and validated using twenty-four groups of homologous proteins. Using HOMA, homology models were generated for 510 proteins, including 264 proteins modeled with correct folds and 246 modeled with incorrect folds. Accuracies of these models were assessed by superimposition on the corresponding experimentally determined structures. A subset of these results was compared with parallel studies of modeling accuracy using several other automated homology modeling approaches. Overall, HOMA provides prediction accuracies similar to other state-of-the-art homology modeling methods. We also provide an evaluation of several structure quality validation tools in assessing the accuracy of homology models generated with HOMA. This study demonstrates that Verify3D (Luthy et al., Nature 1992;356:83-85) and ProsaII (Sippl, Proteins 1993;17:355-362) are most sensitive in distinguishing between homology models with correct or incorrect folds. For homology models that have the correct fold, the steric conformational energy (including primarily the Van der Waals energy), MolProbity clashscore (Word et al., Protein Sci 2000;9:2251-2259), and the PROCHECK G-factors (Laskowski et al., J Biomol NMR 1996;8:477-486) provide sensitive and consistent methods for assessing accuracy and can distinguish between homology models of higher and lower accuracy. As demonstrated in the accompanying paper (Bhattacharya et al., accompanying paper), combinations of these scores for models generated with HOMA provide a basis for distinguishing low from high accuracy models.     

A method for the improvement of threading-based protein models

A new method for the homology-based modeling of protein three-dimensional structures is proposed and evaluated. The alignment of a query sequence to a structural template produced by threading algorithms usually produces low-resolution molecular models. The proposed method attempts to improve these models. In the first stage, a high-coordination lattice approximation of the query protein fold is built by suitable tracking of the incomplete alignment of the structural template and connection of the alignment gaps. These initial lattice folds are very similar to the structures resulting from standard molecular modeling protocols. Then, a Monte Carlo simulated annealing procedure is used to refine the initial structure. The process is controlled by the model's internal force field and a set of loosely defined restraints that keep the lattice chain in the vicinity of the template conformation. The internal force field consists of several knowledge-based statistical potentials that are enhanced by a proper analysis of multiple sequence alignments. The template restraints are implemented such that the model chain can slide along the template structure or even ignore a substantial fraction of the initial alignment. The resulting lattice models are, in most cases, closer (sometimes much closer) to the target structure than the initial threading-based models. All atom models could easily be built from the lattice chains. The method is illustrated on 12 examples of target/template pairs whose initial threading alignments are of varying quality. Possible applications of the proposed method for use in protein function annotation are briefly discussed.     

MODBASE, a database of annotated comparative protein structure models

MODBASE (http://guitar.rockefeller.edu/modbase) is a relational database of annotated comparative protein structure models for all available protein sequences matched to at least one known protein structure. The models are calculated by MODPIPE, an automated modeling pipeline that relies on PSI-BLAST, IMPALA and MODELLER. MODBASE uses the MySQL relational database management system for flexible and efficient querying, and the MODVIEW Netscape plugin for viewing and manipulating multiple sequences and structures. It is updated regularly to reflect the growth of the protein sequence and structure databases, as well as improvements in the software for calculating the models. For ease of access, MODBASE is organized into different datasets. The largest dataset contains models for domains in 304 517 out of 539 171 unique protein sequences in the complete TrEMBL database (23 March 2001); only models based on significant alignments (PSI-BLAST E-value < 10^–4) and models assessed to have the correct fold are included. Other datasets include models for target selection and structure-based annotation by the New York Structural Genomics Research Consortium, models for prediction of genes in the Drosophila melanogaster genome, models for structure determination of several ribosomal particles and models calculated by the MODWEB comparative modeling web server.