Geometrical Conversion of the EGFR Extracellular Domain by Adiabatic Mapping Combining Normal Mode Analysis of the Elastic Network Model and Energy Optimization.

The activation of epidermal growth factor receptor (EGFR) involves the geometrical conversion of the extracellular domain (ECD) from the tethered to the extended forms with the dynamic rearrangement of the relative positions of four subdomains (SDs); however, this conversion process has not yet been thoroughly understood. We compare the two different forms of the X-ray crystal structures of ECD and simulate the ECD conversion process using adiabatic mapping that combines normal mode analysis of the elastic network model (ENM-NMA) and energy optimization. A comparison of the crystal structures reveals the rigidity of the intradomain geometry of the SD-I and -III backbone regardless of the form. The forward mapping from the tethered to the extended forms retains the intradomain geometry of the SD-I and -III backbone and reveals the trends to rearrange the relative positions of SD-I and -III and to dissociate the C-terminal tail of SD-IV from the hairpin loop in SD-II. The reverse mapping from the extended to the tethered forms complements the promotion of ECD conversion in the presence of epidermal growth factor (EGF).

Comprehensive analysis of the Co-structures of dipeptidyl peptidase IV and its inhibitor.

BACKGROUND: We comprehensively analyzed X-ray cocrystal structures of dipeptidyl peptidase IV (DPP-4) and its inhibitor to clarify whether DPP-4 alters its general or partial structure according to the inhibitor used and whether DPP-4 has a common rule for inhibitor binding. RESULTS: All the main and side chains in the inhibitor binding area were minimally altered, except for a few side chains, despite binding to inhibitors of various shapes. Some residues (Arg125, Glu205, Glu206, Tyr662 and Asn710) in the area had binding modes to fix a specific atom of inhibitor to a particular spatial position in DPP-4. We found two specific water molecules that were common to 92 DPP-4 structures. The two water molecules were close to many inhibitors, and seemed to play two roles: maintaining the orientation of the Glu205 and Glu206 side chains through a network via the water molecules, and arranging the inhibitor appropriately at the S2 subsite. CONCLUSIONS: Our study based on high-quality resources may provide a necessary minimum consensus to help in the discovery of a novel DPP-4 inhibitor that is commercially useful.

Quality assessment methods for 3D protein structure models based on a residue-residue distance matrix prediction.

In the absence of experimentally determined three dimensional (3D) structures of proteins, the prediction of protein structures using computational methods is a standard alternative approach in bioinformatics. When using the predicted protein models to compute the native structure of an unknown target protein, estimating the actual quality of the protein models is important for selecting the best or near-best model. Moreover, estimates of the differences between the protein models and the native protein structure are obviously useful to end users who can then decide on the utility of the models for their specific problems. This article describes two new single-model quality assessment (QA) programs, pure single-model QA method (psQA) and a template based QA method (tbQA), that we developed. psQA is a pure single-model QA program that uses a neural network method to predict residue-residue distance matrices of the native protein structures. tbQA is a quasi-single-model QA program that mainly uses target-template sequence alignments and template structures. The performance of these two model QA programs was analyzed in a data set of 24022 models for 94 targets from the 10th critical assessment of protein structure prediction (CASP10) experiment.

Community-wide evaluation of methods for predicting the effect of mutations on protein-protein interactions.

Community-wide blind prediction experiments such as CAPRI and CASP provide an objective measure of the current state of predictive methodology. Here we describe a community-wide assessment of methods to predict the effects of mutations on protein-protein interactions. Twenty-two groups predicted the effects of comprehensive saturation mutagenesis for two designed influenza hemagglutinin binders and the results were compared with experimental yeast display enrichment data obtained using deep sequencing. The most successful methods explicitly considered the effects of mutation on monomer stability in addition to binding affinity, carried out explicit side-chain sampling and backbone relaxation, evaluated packing, electrostatic, and solvation effects, and correctly identified around a third of the beneficial mutations. Much room for improvement remains for even the best techniques, and large-scale fitness landscapes should continue to provide an excellent test bed for continued evaluation of both existing and new prediction methodologies.

United3D: a protein model quality assessment program that uses two consensus based methods.

In protein structure prediction, such as template-based modeling and free modeling (ab initio modeling), the step that assesses the quality of protein models is very important. We have developed a model quality assessment (QA) program United3D that uses an optimized clustering method and a simple Cα atom contact-based potential. United3D automatically estimates the quality scores (Qscore) of predicted protein models that are highly correlated with the actual quality (GDT_TS). The performance of United3D was tested in the ninth Critical Assessment of protein Structure Prediction (CASP9) experiment. In CASP9, United3D showed the lowest average loss of GDT_TS (5.3) among the QA methods participated in CASP9. This result indicates that the performance of United3D to identify the high quality models from the models predicted by CASP9 servers on 116 targets was best among the QA methods that were tested in CASP9. United3D also produced high average Pearson correlation coefficients (0.93) and acceptable Kendall rank correlation coefficients (0.68) between the Qscore and GDT_TS. This performance was competitive with the other top ranked QA methods that were tested in CASP9. These results indicate that United3D is a useful tool for selecting high quality models from many candidate model structures provided by various modeling methods. United3D will improve the accuracy of protein structure prediction.

Method for predicting homology modeling accuracy from amino acid sequence alignment: the power function.

We have devised a power function (PF) that can predict the accuracy of a three-dimensional (3D) structure model of a protein using only amino acid sequence alignments. This Power Function (PF) consists of three parts; (1) the length of a model, (2) a homology identity percent value and (3) the agreement rate between PSI-PRED secondary structure prediction and the secondary structure judgment of a reference protein. The PF value is mathematically computed from the execution process of homology search tools, such as FASTA or various BLAST programs, to obtain the amino acid sequence alignments. There is a high correlation between the global distance test-total score (GDT_TS) value of the protein model based upon the PF score and the GDT_TS(MAX) value used as an index of protein modeling accuracy in the international contest Critical Assessment of Techniques for Protein Structure Prediction (CASP). Accordingly, the PF method is valuable for constructing a highly accurate model without wasteful calculations of homology modeling that is normally performed by an iterative method to move the main chain and side chains in the modeling process. Moreover, a model with higher accuracy can be obtained by combining the models ordered by the PF score with models sorted by the size of the CIRCLE score. The CIRCLE software is a 3D-1D program, in which energetic stabilization is estimated based upon the experimental environment of each amino acid residue in the protein solution or protein crystals.

HUMAN FAMSD-BASE: high quality protein structure model database for the human genome using the FAMSD homology modeling method.

Almost all proteins express their biological functions through the structural conformation of their specific amino acid sequences. Therefore, acquiring the three-dimensional structures of proteins is very important to elucidate the role of a particular protein. We had built protein structure model databases, which is called RIKEN FAMSBASE (http://famshelp.gsc.riken.jp/famsbase/). The RIKEN FAMSBASE is a genome-wide protein structure model database that contains a large number of protein models from many organisms. The HUMAN FAMSBASE that is one part of the RIKEN FAMSBASE contains many protein models for human genes, which are significant in the pharmaceutical and medicinal fields. We have now implemented an update of the human protein modeling database consisting of 242918 constructed models against the number of 20743 human protein sequences with an improved modeling method called Full Automatic protein Modeling System Developed (FAMSD). The results of our benchmark test of the FAMSD method indicated that it has an excellent capability to pack amino acid side-chains with correct torsion angles in addition to the main-chain, while avoiding the formation of atom-atom collisions that are not found in experimental structures. This new protein structure model database for human genes, which is named HUMAN FAMSD-BASE, is open to the public as a component part of the RIKEN FAMSBASE at http://mammalia.gsc.riken.jp/human_famsd/. A significant improvement of the HUMAN FAMSD-BASE in comparison with the preceding HUMAN FAMSBASE was verified in the benchmark test of this paper. The HUMAN FAMSD-BASE will have an important impact on the progress of biological science.

New protein structure model evaluation methods that include a side-chain consensus score for the protein modeling.

Selecting the best quality model from a set of predicted structures is one of the most important aspects of protein structure prediction. We have developed model quality assessment programs that select high quality models which account for both the Calpha backbone and side-chain atom positions. The new methods are based on the consensus method with consideration of the side-chain environment of a protein structure and the secondary structure agreement. This Side-chain Environment Consensus (SEC) method is compared with the conventional consensus method, 3D-Jury (Ginalski K. et al., Bioinformatics, 19, 1015-1018 (2003)), which takes into account only the Calpha backbone atoms of the protein model. As the result, it was found that the SEC method selects the models with more accurate positioning of the side-chain atoms than the 3D-Jury method. When the SEC method was used in combination with the 3D-Jury method (3DJ+SEC), models were selected with improved quality both in the Calpha backbone and side-chain atom positions. Moreover, the CIRCLE (CCL) method (Terashi G. et al., Proteins, 69 (Suppl. 8), 98-107 (2007)) based on the 3D-1D profile score has been shown to select the best possible models that are the closest to the native structures from candidate models. Accordingly, the 3DJ+SEC+CCL method, in which CIRCLE is used after reducing the number of candidates by the 3DJ+SEC consensus method, was found to be very effective in selecting high quality models. Thus, the best method (the 3DJ+SEC+CCL method) includes the consensus approaches of the Calpha backbone and the side-chains, the secondary structure agreement and the 3D-1D profile score which corresponds to the free energy-like score in the residues of the protein model. In short, new algorithms are introduced in protein structure evaluation methods that are based on a side-chain consensus score. Additionally, in order to apply the 3DJ+SEC+CCL method and indicate the usefulness of this method, a model of human Cabin1, a protein associated with p53 function and cancer, is created using various internet modeling and alignment servers.

FAMSD: A powerful protein modeling platform that combines alignment methods, homology modeling, 3D structure quality estimation and molecular dynamics.

The prediction of a protein three-dimensional (3D) structure is one of the most important challenges in computational structural biology. We have developed an automatic protein 3D structure prediction method called FAMSD. FAMSD is based on a comparative modeling method which consists of the following four steps: (1) generating and selecting sequence alignments between target and template proteins; (2) constructing 3D structure models based on each selected alignment; (3) selecting the best 3D structure model and (4) refining the selected model. In the FAMSD method, sequence alignment programs such as a series of BLAST programs, SP3 and SPARKS2 programs, the homology modeling program FAMS (Full Automatic Modeling System), the model quality estimation program CIRCLE and the molecular dynamics program APRICOT were used in combination to construct high quality protein models. To assess the FAMSD method we have participated in the 8th Critical Assessment of Techniques for Protein Structure Prediction (CASP8) experiment. The results of our original assessment indicate that the FAMSD method offers excellent capability in packing side-chains with the correct torsion angles while avoiding the formation of atom-atom collisions. Since side-chain packing plays a significant role in defining the biological function of proteins, this method is a valuable resource in biological, pharmaceutical and medicinal research efforts.

Dynamic influence of the two membrane-proximal immunoglobulin-like domains upon the peptide-binding platform domain in class I and class II major histocompatibility complexes: normal mode analysis.

Major histocompatibility complexes (MHCs) mainly fall into class I and class II. The two classes have similar structures, with two membrane-proximal immunoglobulin-like domains and a peptide-binding platform domain, though their organizations are different. We simulated the dynamics of a whole and partial model deficient in either of the two membrane-proximal domains for class I and class II using normal mode analysis. Our study showed that the influence of the two membrane-proximal domains upon the dynamics of the platform domain were decisively different between class II and class I. Both membrane-proximal domains (the alpha2 and beta2 domains) of class II MHC, especially the alpha2 domain, influenced the most important pocket that accommodates a large hydrophobic anchor side chain of the N-terminal side of the bound peptide, though the pocket was not in the alpha2 domain neighborhood. By contrast, the two membrane-proximal domains (the alpha3 and beta2m domains) of class I MHC had little influence upon the most important pocket that accommodates the N-terminal residue of the bound peptide. These results suggest that the two membrane-proximal domains of class II MHC have a greater influence upon peptide-binding than those of class I MHC.

The SKE-DOCK server and human teams based on a combined method of shape complementarity and free energy estimation.

We participated in rounds 6-12 of the critical assessment of predicted interaction (CAPRI) contest as the SKE-DOCK server and human teams. The SKE-DOCK server is based on simple geometry docking and a knowledge base scoring function. The procedure is summarized in the following three steps: (1) protein docking according to shape complementarity, (2) evaluating complex models, and (3) repacking side-chain of models. The SKE-DOCK server did not make use of biological information. On the other hand, the human team tried various intervention approaches. In this article, we describe in detail the processes of the SKE-DOCK server, together with results and reasons for success and failure. Good predicted models were obtained for target 25 by both the SKE-DOCK server and human teams. When the modeled receptor proteins were superimposed on the experimental structures, the smallest Ligand-rmsd values corresponding to the rmsd between the model and experimental structures were 3.307 and 3.324 A, respectively. Moreover, the two teams obtained 4 and 2 acceptable models for target 25. The overall result for both the SKE-DOCK server and human teams was medium accuracy for one (Target 25) out of nine targets.

Fams-ace: a combined method to select the best model after remodeling all server models.

During Critical Assessment of Protein Structure Prediction (CASP7, Pacific Grove, CA, 2006), fams-ace was entered in the 3D coordinate prediction category as a human expert group. The procedure can be summarized by the following three steps. (1) All the server models were refined and rebuilt utilizing our homology modeling method. (2) Representative structures were selected from each server, according to a model quality evaluation, based on a 3D1D profile score (like Verify3D). (3) The top five models were selected and submitted in the order of the consensus-based score (like 3D-Jury). Fams-ace is a fully automated server and does not require human intervention. In this article, we introduce the methodology of fams-ace and discuss the successes and failures of this approach during CASP7. In addition, we discuss possible improvements for the next CASP.

Residue-residue interactions regulating the Ca2+-induced EF-hand conformation changes in calmodulin.

Calmodulin (CaM) is a Ca2+-binding messenger protein having four Ca2+-binding motifs named 'EF-hand'; the EF-hand motifs undergo a conformation change induced by Ca2+-binding. In order to study how Ca2+-binding induces the conformation change of EF-hand motifs and which residues are involved in the reaction, two 1µ second long MD simulations were independently performed from the apo- and holo-CaM and their structures and interactions were compared. The Ca2+-binding weakens the helix-helix interaction in all EF-hand, however, the holo-CaM MD adopted the close-like form. The correlation coefficients obtained from the two MDs show the residues comprising interactions being involved in their close-open conformation changes; most of these residues are hydrophobic amino acids but some of them are hydrophilic (T34, H107, N111 and Q143). The hydrophilic residues are expected to lock the EF-hands by their side-chains and main-chain carbonyl oxygen of another hydrophobic residue. Furthermore, the interaction pattern of EF-hand3 and 4 are similar to each other. On the other hand, the interaction pattern of EF-hand2 is different from others; its polar residues are expected to play an important role in regulating the EF-hand2 conformation.

FAMS complex: a fully automated homology modeling system for protein complex structures.

The formation of a protein-protein complex is responsible for many biological functions; therefore, three-dimensional structures of protein complexes are essential for deeper understandings of protein functions and the mechanisms of diseases at the atomic level. However, compared with individual proteins, complex structures are difficult to solve experimentally because of technical limitations. Thus a method that can predict protein complex structures would be invaluable. In this study, we developed new software, FAMS Complex; a fully automated homology modeling system for protein complex structures consisting of two or more molecules. FAMS Complex requires only sequences and alignments of the target protein as input and constructs all molecules simultaneously and automatically. FAMS Complex is likely to become an essential tool for structure-based drug design, such as in silico screening to accelerate drug discovery before an experimental structure is solved. Moreover, in this post-genomic era when huge amounts of protein sequence information are available, a major goal is the determination of protein-protein interaction networks on a genomic scale. FAMS Complex will contribute to this goal, because its procedure is fully automated and so is suited for large-scale genome wide modeling.

Searching for protein-protein interaction sites and docking by the methods of molecular dynamics, grid scoring, and the pairwise interaction potential of amino acid residues.

In CAPRI Rounds 1 and 2, we assumed that because there are many ionic charges that weaken electrostatic interaction forces in living cells, the hydrophobic interaction force might be important entropically. As a result of Rounds 1 and 2, the predictions for binding sites and geometric centers were acceptable, but those of the binding axes were poor, because only the largest benzene cluster was used for generating the initial docking structures. These were generated by fitting of benzene clusters formed on the surface of receptor and ligand. In CAPRI Rounds 3-5, the grid-scoring sum on the protein-protein interaction surface and the pairwise potential of the amino acid residues, which were indicated as coming easily into the protein-protein interaction regions, were used as the calculation methods, along with the smaller benzene clusters that participated in benzene cluster fitting. Good predicted models were obtained for Targets 11 and 12. When the modeled receptor proteins were superimposed on the experimental structures, the smallest ligand root-mean-square deviation (RMSD) values corresponding to the RMSD between the model and experimental structures were 6.2 A and 7.3 A, respectively.

Protein structure prediction in CASP6 using CHIMERA and FAMS.

In CASP6, the CHIMERA-group predicted full-atom models of all targets using SKE-CHIMERA, a Web-user interface system for protein structure prediction that allows human intervention at necessary stages; we used a lot of information from our own data and from publicly available data. Using SKE-CHIMERA, we iterated manual step (template selection and alignment by the in-house program CHIMERA) and automatic step (three-dimensional model building by the in-house program FAMS). The official CASP6 assessment showed that CHIMERA-group was one of the most successful predictors in homology modeling, especially for FR/H (Fold Recognition/Homologous). In this article, we introduce the method of CHIMERA-group and discuss its successes and failures in CASP6.

CAB-Align: A Flexible Protein Structure Alignment Method Based on the Residue-Residue Contact Area.

Proteins are flexible, and this flexibility has an essential functional role. Flexibility can be observed in loop regions, rearrangements between secondary structure elements, and conformational changes between entire domains. However, most protein structure alignment methods treat protein structures as rigid bodies. Thus, these methods fail to identify the equivalences of residue pairs in regions with flexibility. In this study, we considered that the evolutionary relationship between proteins corresponds directly to the residue-residue physical contacts rather than the three-dimensional (3D) coordinates of proteins. Thus, we developed a new protein structure alignment method, contact area-based alignment (CAB-align), which uses the residue-residue contact area to identify regions of similarity. The main purpose of CAB-align is to identify homologous relationships at the residue level between related protein structures. The CAB-align procedure comprises two main steps: First, a rigid-body alignment method based on local and global 3D structure superposition is employed to generate a sufficient number of initial alignments. Then, iterative dynamic programming is executed to find the optimal alignment. We evaluated the performance and advantages of CAB-align based on four main points: (1) agreement with the gold standard alignment, (2) alignment quality based on an evolutionary relationship without 3D coordinate superposition, (3) consistency of the multiple alignments, and (4) classification agreement with the gold standard classification. Comparisons of CAB-align with other state-of-the-art protein structure alignment methods (TM-align, FATCAT, and DaliLite) using our benchmark dataset showed that CAB-align performed robustly in obtaining high-quality alignments and generating consistent multiple alignments with high coverage and accuracy rates, and it performed extremely well when discriminating between homologous and nonhomologous pairs of proteins in both single and multi-domain comparisons. The CAB-align software is freely available to academic users as stand-alone software at http://www.pharm.kitasato-u.ac.jp/bmd/bmd/Publications.html.

Interaction between the antigen and antibody is controlled by the constant domains: normal mode dynamics of the HEL-HyHEL-10 complex.

The antigen binding fragment (Fab) of a monoclonal antibody (HyHEL-10) consists of variable domains (Fv) and constant domains (CL-CH1). Normal modes have been calculated from the three-dimensional structures of hen egg lysozyme (HEL) with Fab, those of HEL with Fv, and so on. Only a small structural change was found between HEL-Fab and HEL-Fv complexes. However, HEL-Fv had a one order of magnitude lower dissociation constant than HEL-Fab. The Calpha fluctuations of HEL-Fab differed from those of HEL-Fv with normal mode calculation, and the dynamics can be thought to be related to the protein-protein interactions. CL-CH1 may have influence not only around local interfaces between CL-CH1 and Fv, but also around the interacting regions between HEL and Fv, which are longitudinally distant. Eighteen water molecules were found in HEL-Fv around the interface between HEL and Fv compared with one water molecule in HEL-Fab. These solvent molecules may occupy the holes and channels, which may occur due to imperfect complementarity of the complex. Therefore, the suppression of atomic vibration around the interface between Fv and HEL can be thought to be related to favorable and compact interface formation by complete desolvation. It is suggested that the ability to control the antigen-antibody affinity is obtained from modifying the CL-CH1. The second upper loop in the constant domain of the light chain (UL2-CL), which is a conserved gene in several light chains, showed the most remarkable fluctuation changes. UL2-CL could play an important role and could be attractive for modification in protein engineering.

Evaluation of the third solvent clusters fitting procedure for the prediction of protein-protein interactions based on the results at the CAPRI blind docking study.

To predict protein-protein interactions, rough or coarse handling for the induced fit problem is proposed. Our method involves the overlap of two hydrophobic interactions as "third solvent clusters fitting." Predictions for binding sites and geometric centers were acceptable, but those of the binding axes were poor. In this study, only the largest benzene cluster was used for the third solvent clusters fitting. For the next CAPRI targets, we must perform protein-protein interaction analyses, which include other smaller benzene clusters.

Protein structure prediction in structure based drug design.

The human genome and other genome sequencing projects have generated huge amounts of protein sequence information. Recently, a structural genomics project that aims to determine at least one representative three-dimensional structure from every protein family experimentally has been started. Homology modeling will play an essential role in structure based drug design such as in silico screening; because based on these representative structures the three-dimensional structures of the remaining proteins encoded in the various genomes can be predicted by homology modeling. The results of the last Critical Assessment of Techniques for Protein Structure Prediction (CASP5) demonstrated that the quality of homology modeling prediction has improved; reaching a level of reliability that biologists can now confidently use homology modeling. With improvements in modeling software and the growing number of known protein structures, homology modeling is becoming a more and more powerful and reliable tool. The present paper discusses the features and roles of homology modeling in structure based drug design, and describes the CHIMERA and FAMS modeling systems as examples. For a sample application, homology modeling of non-structural proteins of the severe acute respiratory syndrome (SARS) coronavirus is discussed. Many biological functions involve formation of protein-protein complexes; in which the protein molecules behave dynamically in the course of binding. Therefore, an understanding of protein-protein interaction will be very important for structure based drug design. To this end, normal mode analysis is useful. The present paper discusses the prediction of protein-protein interaction using normal mode analysis and examples of applications are given.