Generalized comparative modeling (GENECOMP): a combination of sequence comparison, threading, and lattice modeling for protein structure prediction and refinement

Modeling protein structure as a stable static equilibrium

We present evidence that the protein structure can be modeled as a stable static equilibrium, determined mainly by compressive supports in the nonpolar interior. That is, protein structures derive their structural strength through the same mechanical principles as do conventional structures like bridges and buildings. This is based on the observation that the experimentally elucidated structural determinants, the interior nonpolar side chains, are engaged in strong compressions in static terms. At the same time, major substructures in proteins, helices and h-bonded strands, because of their geometry, inherently leave gaps in the space they occupy. Under the compressive force, nonpolar side chains from one substructure can protrude into the gaps of another neighboring substructure and block its motion. As a result, interlocking of substructures can form, which builds up the nonpolar core assembly. The native structure then is the one with the structurally most stable core assembly. While intuitively appealing, this is a radical departure from the prevailing thinking that protein native structure is determined by global energy minimum, which is founded on thermodynamic hypothesis. Furthermore, to develop an effective model for analyzing protein structure with conventional tools, a proper mechanical representation must be established. By proving that the stability of the equilibrium in compressive interactions is conditioned on a form of mechanical energy minimum, we show that our notion of native structure can be equally consistent with the thermodynamic hypothesis. By mathematically treating the blocking action, an interaction, as a bar, a physical object, we succeed in representing and analyzing the core assembly as truss, a conventional structure. In this paper we define and expound step-by-step increasingly integrated interlocking patterns. We then analyze the core assemblies of a large set of diverse protein database structures. A native structure can be distinguished from decoys by comparing the composition and strength of their core assemblies. We show the results for two sets of native structures vs decoys.

Homology modeling of human histone deacetylase 10 and design of potential selective inhibitors

Histone deacetylases (HDACs) are implicated in the pathology of various cancers, and their pharmacological blockade has proven to be promising in reversing the malignant phenotypes. However, lack of crystal structures of some of the human HDAC isoforms (e.g., HDAC10) hinders the design of the isoform-selective inhibitor. Here, the recently solved X-ray crystal structure of Danio rerio (zebrafish) HDAC10 (Protein Data Bank (PDB) ID; 5TD7, released on 24 May 2017) was retrieved from the PDB and used as a template structure to model the three-dimensional structure of human HDAC10. The overall quality of the best model (M0017) was assessed by computing its z-score-a measure of the deviation of the total energy of the structure with respect to an energy distribution derived from random conformations and by docking of known HDAC10 inhibitors to its catalytic cavity. Furthermore, to identify potential HDAC10-selective inhibitor ligand-based virtual screening was carried out against the ZINC database. The free modeled structure of HDAC10 and its complexes with quisinostat and the highest-ranked compound ZINC19749069 were submitted to molecular dynamics simulation. The comparative analysis of root-mean-squared deviation, root-mean-squared fluctuation, radius of gyration (Rg), and potential energy of these systems showed that HDAC10-ZINC19749069 complex remained the most stable over time. Thus, M0017 could be potentially used for structure-based inhibitor against HDAC10, and ZINC19749069 may provide a scaffold for further optimization. Communicated by Ramaswamy H. Sarma.

SURPASS Low-Resolution Coarse-Grained Protein Modeling

Coarse-grained modeling of biomolecules has a very important role in molecular biology. In this work we present a novel SURPASS (Single United Residue per Pre-Averaged Secondary Structure fragment) model of proteins that can be an interesting alternative for existing coarse-grained models. The design of the model is unique and strongly supported by the statistical analysis of structural regularities characteristic for protein systems. Coarse-graining of protein chain structures assumes a single center of interactions per residue and accounts for preaveraged effects of four adjacent residue fragments. Knowledge-based statistical potentials encode complex interaction patterns of these fragments. Using the Replica Exchange Monte Carlo sampling scheme and a generic version of the SURPASS force field we performed test simulations of a representative set of single-domain globular proteins. The method samples a significant part of conformational space and reproduces protein structures, including native-like, with surprisingly good accuracy. Future extension of the SURPASS model on large biomacromolecular systems is briefly discussed.

One-Dimensional Structural Properties of Proteins in the Coarse-Grained CABS Model

Despite the significant increase in computational power, molecular modeling of protein structure using classical all-atom approaches remains inefficient, at least for most of the protein targets in the focus of biomedical research. Perhaps the most successful strategy to overcome the inefficiency problem is multiscale modeling to merge all-atom and coarse-grained models. This chapter describes a well-established CABS coarse-grained protein model. The CABS (C-Alpha, C-Beta, and Side chains) model assumes a 2-4 united-atom representation of amino acids, knowledge-based force field (derived from the statistical regularities seen in known protein sequences and structures) and efficient Monte Carlo sampling schemes (MC dynamics, MC replica-exchange, and combinations). A particular emphasis is given to the unique design of the CABS force-field, which is largely defined using one-dimensional structural properties of proteins, including protein secondary structure. This chapter also presents CABS-based modeling methods, including multiscale tools for de novo structure prediction, modeling of protein dynamics and prediction of protein-peptide complexes. CABS-based tools are freely available at http://biocomp.chem.uw.edu.pl/tools.

Computational methods for constructing protein structure models from 3D electron microscopy maps

Protein structure determination by cryo-electron microscopy (EM) has made significant progress in the past decades. Resolutions of EM maps have been improving as evidenced by recently reported structures that are solved at high resolutions close to 3Å. Computational methods play a key role in interpreting EM data. Among many computational procedures applied to an EM map to obtain protein structure information, in this article we focus on reviewing computational methods that model protein three-dimensional (3D) structures from a 3D EM density map that is constructed from two-dimensional (2D) maps. The computational methods we discuss range from de novo methods, which identify structural elements in an EM map, to structure fitting methods, where known high resolution structures are fit into a low-resolution EM map. A list of available computational tools is also provided.

Multi-LZerD: multiple protein docking for asymmetric complexes

The tertiary structures of protein complexes provide a crucial insight about the molecular mechanisms that regulate their functions and assembly. However, solving protein complex structures by experimental methods is often more difficult than single protein structures. Here, we have developed a novel computational multiple protein docking algorithm, Multi-LZerD, that builds models of multimeric complexes by effectively reusing pairwise docking predictions of component proteins. A genetic algorithm is applied to explore the conformational space followed by a structure refinement procedure. Benchmark on eleven hetero-multimeric complexes resulted in near-native conformations for all but one of them (a root mean square deviation smaller than 2.5Å). We also show that our method copes with unbound docking cases well, outperforming the methodology that can be directly compared with our approach. Multi-LZerD was able to predict near-native structures for multimeric complexes of various topologies.

PROTEIN THREADING WITH PROFILES AND DISTANCE CONSTRAINTS USING CLIQUE BASED ALGORITHMS

With the advent of experimental technologies like chemical cross-linking, it has become possible to obtain distances between specific residues of a newly sequenced protein. These types of experiments usually are less time consuming than X-ray crystallography or NMR. Consequently, it is highly desired to develop a method that incorporates this distance information to improve the performance of protein threading methods. However, protein threading with profiles in which constraints on distances between residues are given is known to be NP-hard. By using the notion of a maximum edge-weight clique finding algorithm, we introduce a more efficient method called FTHREAD for profile threading with distance constraints that is 18 times faster than its predecessor CLIQUETHREAD. Moreover, we also present a novel practical algorithm NTHREAD for profile threading with Non-strict constraints. The overall performance of FTHREAD on a data set shows that although our algorithm uses a simple threading function, our algorithm performs equally well as some of the existing methods. Particularly, when there are some unsatisfied constraints, NTHREAD (Non-strict constraints threading algorithm) performs better than threading with FTHREAD (Strict constraints threading algorithm). We have also analyzed the effects of using a number of distance constraints. This algorithm helps the enhancement of alignment quality between the query sequence and template structure, once the corresponding template structure is determined for the target sequence.

Studies on the rules of β-strand alignment in a protein β-sheet structure

To further disclose the underlying mechanisms of protein β-sheet formation, studies were made on the rules of β-strands alignment forming β-sheet structure using statistical and machine learning approaches. Firstly, statistical analysis was performed on the sum of β-strands between each β-strand pairs in protein sequences. The results showed a propensity of near-neighbor pairing (or called "first come first pair") in the β-strand pairs. Secondly, based on the same dataset, the pairwise cross-combinations of real β-strand pairs and four pseudo-β-strand contained pairs were classified by support vector machine (SVM). A novel feature extracting approach was designed for classification using the average amino acid pairing encoding matrix (APEM). Analytical results of the classification indicated that a segment of β-strand had the ability to distinguish β-strands from segments of α-helix and coil. However, the result also showed that a β-strand was not strongly conserved to choose its real partner from all the alternative β-strand partners, which was corresponding with the ordination results of the statistical analysis each other. Thus, the rules of "first come first pair" propensity and the non-conservative ability to choose real partner, were possible important factors affecting the β-strands alignment forming β-sheet structures.

Effect of using suboptimal alignments in template-based protein structure prediction

Computational protein structure prediction remains a challenging task in protein bioinformatics. In the recent years, the importance of template-based structure prediction is increasing because of the growing number of protein structures solved by the structural genomics projects. To capitalize the significant efforts and investments paid on the structural genomics projects, it is urgent to establish effective ways to use the solved structures as templates by developing methods for exploiting remotely related proteins that cannot be simply identified by homology. In this work, we examine the effect of using suboptimal alignments in template-based protein structure prediction. We showed that suboptimal alignments are often more accurate than the optimal one, and such accurate suboptimal alignments can occur even at a very low rank of the alignment score. Suboptimal alignments contain a significant number of correct amino acid residue contacts. Moreover, suboptimal alignments can improve template-based models when used as input to Modeller. Finally, we use suboptimal alignments for handling a contact potential in a probabilistic way in a threading program, SUPRB. The probabilistic contacts strategy outperforms the partly thawed approach, which only uses the optimal alignment in defining residue contacts, and also the re-ranking strategy, which uses the contact potential in re-ranking alignments. The comparison with existing methods in the template-recognition test shows that SUPRB is very competitive and outperforms existing methods.

The interstrand amino acid pairs play a significant role in determining the parallel or antiparallel orientation of beta-strands

It is widely considered that it is not appropriate to treat beta-pairs in isolation, since other secondary structural models (such as helices, coils), protein topology and protein tertiary structures would limit beta-strand pairing. However, to understand the underlying mechanisms of beta-sheet formation, studies ought to be performed separately on more concrete aspects. In this study, we focus on the parallel or antiparallel orientation of beta-strands. First, statistical analysis was performed on the relative frequencies of the interstrand amino acid pairs within parallel and antiparallel beta-strands. Consequently, features were extracted by singular value decomposition from the statistical results. By using the support vector machine to distinguish the features extracted from the two types of beta-strands, high accuracy was achieved (up to 99.4%). This suggests that the interstrand amino acid pairs play a significant role in determining the parallel or antiparallel orientation of beta-strands. These results may provide useful information for developing other useful algorithms to examine to the beta-strand folding pathways, and could eventually lead to protein structure predictions.

Estimating quality of template-based protein models by alignment stability

The error in protein tertiary structure prediction is unavoidable, but it is not explicitly shown in most of the current prediction algorithms. Estimated error of a predicted structure is crucial information for experimental biologists to use the prediction model for design and interpretation of experiments. Here, we propose a method to estimate errors in predicted structures based on the stability of the optimal target-template alignment when compared with a set of suboptimal alignments. The stability of the optimal alignment is quantified by an index named the SuboPtimal Alignment Diversity (SPAD). We implemented SPAD in a profile-based threading algorithm and investigated how well SPAD can indicate errors in threading models using a large benchmark dataset of 5232 alignments. SPAD shows a very good correlation not only to alignment shift errors but also structure-level errors, the root mean square deviation (RMSD) of predicted structure models to the native structures (i.e. global errors), and local errors at each residue position. We have further compared SPAD with seven other quality measures, six from sequence alignment-based measures and one atomic statistical potential, discrete optimized protein energy (DOPE), in terms of the correlation coefficient to the global and local structure-level errors. In terms of the correlation to the RMSD of structure models, when a target and a template are in the same SCOP family, the sequence identity showed a best correlation to the RMSD; in the superfamily level, SPAD was the best; and in the fold level, DOPE was best. However, in a head-to-head comparison, SPAD wins over the other measures. Next, SPAD is compared with three other measures of local errors. In this comparison, SPAD was best in all of the family, the superfamily and the fold levels. Using the discovered correlation, we have also predicted the global and local error of our predicted structures of CASP7 targets by the SPAD. Finally, we proposed a sausage representation of predicted tertiary structures which intuitively indicate the predicted structure and the estimated error range of the structure simultaneously.

Characterization of local geometry of protein surfaces with the visibility criterion

Experimentally determined protein tertiary structures are rapidly accumulating in a database, partly due to the structural genomics projects. Included are proteins of unknown function, whose function has not been investigated by experiments and was not able to be predicted by conventional sequence-based search. Those uncharacterized protein structures highlight the urgent need of computational methods for annotating proteins from tertiary structures, which include function annotation methods through characterizing protein local surfaces. Toward structure-based protein annotation, we have developed VisGrid algorithm that uses the visibility criterion to characterize local geometric features of protein surfaces. Unlike existing methods, which only concerns identifying pockets that could be potential ligand-binding sites in proteins, VisGrid is also aimed to identify large protrusions, hollows, and flat regions, which can characterize geometric features of a protein structure. The visibility used in VisGrid is defined as the fraction of visible directions from a target position on a protein surface. A pocket or a hollow is recognized as a cluster of positions with a small visibility. A large protrusion in a protein structure is recognized as a pocket in the negative image of the structure. VisGrid correctly identified 95.0% of ligand-binding sites as one of the three largest pockets in 5616 benchmark proteins. To examine how natural flexibility of proteins affects pocket identification, VisGrid was tested on distorted structures by molecular dynamics simulation. Sensitivity decreased approximately 20% for structures of a root mean square deviation of 2.0 A to the original crystal structure, but specificity was not much affected. Because of its intuitiveness and simplicity, the visibility criterion will lay the foundation for characterization and function annotation of local shape of proteins.

A multi-template combination algorithm for protein comparative modeling

BACKGROUND

Multiple protein templates are commonly used in manual protein structure prediction. However, few automated algorithms of selecting and combining multiple templates are available.

RESULTS

Here we develop an effective multi-template combination algorithm for protein comparative modeling. The algorithm selects templates according to the similarity significance of the alignments between template and target proteins. It combines the whole template-target alignments whose similarity significance score is close to that of the top template-target alignment within a threshold, whereas it only takes alignment fragments from a less similar template-target alignment that align with a sizable uncovered region of the target. We compare the algorithm with the traditional method of using a single top template on the 45 comparative modeling targets (i.e. easy template-based modeling targets) used in the seventh edition of Critical Assessment of Techniques for Protein Structure Prediction (CASP7). The multi-template combination algorithm improves the GDT-TS scores of predicted models by 6.8% on average. The statistical analysis shows that the improvement is significant (p-value < 10-4). Compared with the ideal approach that always uses the best template, the multi-template approach yields only slightly better performance. During the CASP7 experiment, the preliminary implementation of the multi-template combination algorithm (FOLDpro) was ranked second among 67 servers in the category of high-accuracy structure prediction in terms of GDT-TS measure.

CONCLUSION

We have developed a novel multi-template algorithm to improve protein comparative modeling.

Use of secondary structural information and C alpha-C alpha distance restraints to model protein structures with MODELLER

Protein secondary structure predictions and amino acid long range contact map predictions from primary sequence of proteins have been explored to aid in modelling protein tertiary structures. In order to evaluate the usefulness of secondary structure and 3D-residue contact prediction methods to model protein structures we have used the known Q3 (alpha-helix,beta-strands and irregular turns/loops) secondary structure information, along with residue-residue contact information as restraints for MODELLER. We present here results of our modelling studies on 30 best resolved single domain protein structures of varied lengths. The results shows that it is very difficult to obtain useful models even with 100% accurate secondary structure predictions and accurate residue contact predictions for up to 30% of residues in a sequence. The best models that we obtained for proteins of lengths 37, 70, 118, 136 and 193 amino acid residues are of RMSDs 4.17, 5.27, 9.12, 7.89 and 9.69,respectively. The results show that one can obtain better models for the proteins which have high percent of alpha-helix content. This analysis further shows that MODELLER restrain optimization program can be useful only if we have truly homologous structure(s) as a template where it derives numerous restraints, almost identical to the templates used. This analysis also clearly indicates that even if we satisfy several true residue-residue contact distances, up to 30%of their sequence length with fully known secondary structural information, we end up predicting model structures much distant from their corresponding native structures.

Conservative secondary structure motifs already present in early-stage folding (in silico) as found in serpines family

The well-known procedure implemented in ClustalW oriented on the sequence comparison was applied to structure comparison. The consensus sequence as well as consensus structure has been defined for proteins belonging to serpine family. The structure of early stage intermediate was the object for similarity search. The high values of W(sequence) appeared to be accordant with high values of W(structure) making possible structure comparison using common criteria for sequence and structure comparison. Since the early stage structural form has been created according to limited conformational sub-space which does not include the beta-structure (this structure is mediated by C7eq structural form), is particularly important to see, that the C7eq structural form may be treated as the seed for beta-structure present in the final native structure of protein. The applicability of ClustalW procedure to structure comparison makes these two comparisons unified.

Protein structure prediction: combining de novo modeling with sparse experimental data

Routine structure prediction of new folds is still a challenging task for computational biology. The challenge is not only in the proper determination of overall fold but also in building models of acceptable resolution, useful for modeling the drug interactions and protein-protein complexes. In this work we propose and test a comprehensive approach to protein structure modeling supported by sparse, and relatively easy to obtain, experimental data. We focus on chemical shift-based restraints from NMR, although other sparse restraints could be easily included. In particular, we demonstrate that combining the typical NMR software with artificial intelligence-based prediction of secondary structure enhances significantly the accuracy of the restraints for molecular modeling. The computational procedure is based on the reduced representation approach implemented in the CABS modeling software, which proved to be a versatile tool for protein structure prediction during the CASP (CASP stands for critical assessment of techniques for protein structure prediction) experiments (see http://predictioncenter/CASP6/org). The method is successfully tested on a small set of representative globular proteins of different size and topology, including the two CASP6 targets, for which the required NMR data already exist. The method is implemented in a semi-automated pipeline applicable to a large scale structural annotation of genomic data. Here, we limit the computations to relatively small set. This enabled, without a loss of generality, a detailed discussion of various factors determining accuracy of the proposed approach to the protein structure prediction.

Comparative modeling without implicit sequence alignments

MOTIVATION

The number of known protein sequences is about thousand times larger than the number of experimentally solved 3D structures. For more than half of the protein sequences a close or distant structural analog could be identified. The key starting point in a classical comparative modeling is to generate the best possible sequence alignment with a template or templates. With decreasing sequence similarity, the number of errors in the alignments increases and these errors are the main causes of the decreasing accuracy of the molecular models generated. Here we propose a new approach to comparative modeling, which does not require the implicit alignment - the model building phase explores geometric, evolutionary and physical properties of a template (or templates).

RESULTS

The proposed method requires prior identification of a template, although the initial sequence alignment is ignored. The model is built using a very efficient reduced representation search engine CABS to find the best possible superposition of the query protein onto the template represented as a 3D multi-featured scaffold. The criteria used include: sequence similarity, predicted secondary structure consistency, local geometric features and hydrophobicity profile. For more difficult cases, the new method qualitatively outperforms existing schemes of comparative modeling. The algorithm unifies de novo modeling, 3D threading and sequence-based methods. The main idea is general and could be easily combined with other efficient modeling tools as Rosetta, UNRES and others.

Statistical potential-based amino acid similarity matrices for aligning distantly related protein sequences

Aligning distantly related protein sequences is a long-standing problem in bioinformatics, and a key for successful protein structure prediction. Its importance is increasing recently in the context of structural genomics projects because more and more experimentally solved structures are available as templates for protein structure modeling. Toward this end, recent structure prediction methods employ profile-profile alignments, and various ways of aligning two profiles have been developed. More fundamentally, a better amino acid similarity matrix can improve a profile itself; thereby resulting in more accurate profile-profile alignments. Here we have developed novel amino acid similarity matrices from knowledge-based amino acid contact potentials. Contact potentials are used because the contact propensity to the other amino acids would be one of the most conserved features of each position of a protein structure. The derived amino acid similarity matrices are tested on benchmark alignments at three different levels, namely, the family, the superfamily, and the fold level. Compared to BLOSUM45 and the other existing matrices, the contact potential-based matrices perform comparably in the family level alignments, but clearly outperform in the fold level alignments. The contact potential-based matrices perform even better when suboptimal alignments are considered. Comparing the matrices themselves with each other revealed that the contact potential-based matrices are very different from BLOSUM45 and the other matrices, indicating that they are located in a different basin in the amino acid similarity matrix space.

Prediction of beta-strand packing interactions using the signature product

The prediction of beta-sheet topology requires the consideration of long-range interactions between beta-strands that are not necessarily consecutive in sequence. Since these interactions are difficult to simulate using ab initio methods, we propose a supplementary method able to assign beta-sheet topology using only sequence information. We envision using the results of our method to reduce the three-dimensional search space of ab initio methods. Our method is based on the signature molecular descriptor, which has been used previously to predict protein-protein interactions successfully, and to develop quantitative structure-activity relationships for small organic drugs and peptide inhibitors. Here, we show how the signature descriptor can be used in a Support Vector Machine to predict whether or not two beta-strands will pack adjacently within a protein. We then show how these predictions can be used to order beta-strands within beta-sheets. Using the entire PDB database with ten-fold cross-validation, we have achieved 74.0% accuracy in packing prediction and 75.6% accuracy in the prediction of edge strands. For the case of beta-strand ordering, we are able to predict the correct ordering accurately for 51.3% of the beta-sheets. Furthermore, using a simple confidence metric, we can determine those sheets for which accurate predictions can be obtained. For the top 25% highest confidence predictions, we are able to achieve 95.7% accuracy in beta-strand ordering. [Figure: see text].

Protein-Structure Prediction by Recombination of Fragments

The field of protein-structure prediction has been revolutionized by the application of "mix-and-match" methods both in template-based homology modeling and in template-free de novo folding. Consensus analysis and recombination of fragments copied from known protein structures is currently the only approach that allows the building of models that are closer to the native structure of the target protein than the structure of its closest homologue. It is also the most successful approach in cases in which the target protein exhibits a novel three-dimensional fold. This review summarizes the recent developments in both template-based and template-free protein structure modeling and compares the available methods for protein-structure prediction by recombination of fragments. A convergence between the "protein folding" and "protein evolution" schools of thought is postulated.

A restraint molecular dynamics and simulated annealing approach for protein homology modeling utilizing mean angles

BACKGROUND

We have developed the program PERMOL for semi-automated homology modeling of proteins. It is based on restrained molecular dynamics using a simulated annealing protocol in torsion angle space. As main restraints defining the optimal local geometry of the structure weighted mean dihedral angles and their standard deviations are used which are calculated with an algorithm described earlier by Doker et al. (1999, BBRC, 257, 348-350). The overall long-range contacts are established via a small number of distance restraints between atoms involved in hydrogen bonds and backbone atoms of conserved residues. Employing the restraints generated by PERMOL three-dimensional structures are obtained using standard molecular dynamics programs such as DYANA or CNS.

RESULTS

To test this modeling approach it has been used for predicting the structure of the histidine-containing phosphocarrier protein HPr from E. coli and the structure of the human peroxisome proliferator activated receptor gamma (Ppar gamma). The divergence between the modeled HPr and the previously determined X-ray structure was comparable to the divergence between the X-ray structure and the published NMR structure. The modeled structure of Ppar gamma was also very close to the previously solved X-ray structure with an RMSD of 0.262 nm for the backbone atoms.

CONCLUSION

In summary, we present a new method for homology modeling capable of producing high-quality structure models. An advantage of the method is that it can be used in combination with incomplete NMR data to obtain reasonable structure models in accordance with the experimental data.

PERMOL: restraint-based protein homology modeling using DYANA or CNS

SUMMARY

PERMOL is a new restraint-based program for homology modeling of proteins. Restraints are generated from the information contained in structures of homologous template proteins. Employing the restraints generated by PERMOL, three-dimensional structures are obtained using MD programs such as DYANA or CNS. In contrast to other programs PERMOL is mainly based on the use of dihedral angle information which is optimally suited to preserve the local secondary structure. The global arrangement of these elements is then facilitated by a small number of distance restraints. Using PERMOL homology, models of high quality are obtained. A key advantage of the proposed method is its flexibility, which allows the inclusion of data from other sources, such as experimental restraints and the use of modern molecular dynamics programs to calculate structures.

AVAILABILITY

The software and a detailed manual are available free of charge (http://www.biologie.uni-regensburg.de/Biophysik/Kalbitzer/permol/permol.html)

Protein structure prediction based on fragment assembly and parameter optimization

We propose a novel method for ab-initio prediction of protein tertiary structures based on the fragment assembly and global optimization. Fifteen residue long fragment libraries are constructed using the secondary structure prediction method PREDICT, and fragments in these libraries are assembled to generate full-length chains of a query protein. Tertiary structures of 50 to 100 conformations are obtained by minimizing an energy function for proteins, using the conformational space annealing method that enables one to sample diverse low-lying local minima of the energy. Then in order to enhance the performance of the prediction method, we optimize the linear parameters of the energy function, so that the native-like conformations become energetically more favorable than the non-native ones for proteins with known structures. We test the feasibility of the parameter optimization procedure by applying it to the training set consisting of three proteins: the 10-55 residue fragment of staphylococcal protein A (PDB ID 1bdd), a designed protein betanova, and 1fsd.

Prediction of protein tertiary structure using PROFESY, a novel method based on fragment assembly and conformational space annealing

A novel method for ab initio prediction of protein tertiary structures, PROFESY (PROFile Enumerating SYstem), is proposed. This method utilizes the secondary structure prediction information of a query sequence and the fragment assembly procedure based on global optimization. Fifteen-residue-long fragment libraries are constructed using the secondary structure prediction method PREDICT, and fragments in these libraries are assembled to generate full-length chains of a query protein. Tertiary structures of 50 to 100 conformations are obtained by minimizing an energy function for proteins, using the conformational space annealing method that enables one to sample diverse low-lying local minima of the energy. We apply PROFESY for benchmark tests to proteins with known structures to demonstrate its feasibility. In addition, we participated in CASP5 and applied PROFESY to four new-fold targets for blind prediction. The results are quite promising, despite the fact that PROFESY was in its early stages of development. In particular, PROFESY successfully provided us the best model-one structure for the target T0161.

Protein fragment reconstruction using various modeling techniques

Recently developed reduced models of proteins with knowledge-based force fields have been applied to a specific case of comparative modeling. From twenty high resolution protein structures of various structural classes, significant fragments of their chains have been removed and treated as unknown. The remaining portions of the structures were treated as fixed - i.e., as templates with an exact alignment. Then, the missed fragments were reconstructed using several modeling tools. These included three reduced types of protein models: the lattice SICHO (Side Chain Only) model, the lattice CABS (Calpha + Cbeta + Side group) model and an off-lattice model similar to the CABS model and called REFINER. The obtained reduced models were compared with more standard comparative modeling tools such as MODELLER and the SWISS-MODEL server. The reduced model results are qualitatively better for the higher resolution lattice models, clearly suggesting that these are now mature, competitive and complementary (in the range of sparse alignments) to the classical tools of comparative modeling. Comparison between the various reduced models strongly suggests that the essential ingredient for the sucessful and accurate modeling of protein structures is not the representation of conformational space (lattice, off-lattice, all-atom) but, rather, the specificity of the force fields used and, perhaps, the sampling techniques employed. These conclusions are encouraging for the future application of the fast reduced models in comparative modeling on a genomic scale.

A new approach to prediction of short-range conformational propensities in proteins

MOTIVATION

Knowledge-based potentials are valuable tools for protein structure modeling and evaluation of the quality of the structure prediction obtained by a variety of methods. Potentials of such type could be significantly enhanced by a proper exploitation of the evolutionary information encoded in related protein sequences. The new potentials could be valuable components of threading algorithms, ab-initio protein structure prediction, comparative modeling and structure modeling based on fragmentary experimental data.

RESULTS

A new potential for scoring local protein geometry is designed and evaluated. The approach is based on the similarity of short protein fragments measured by an alignment of their sequence profiles. Sequence specificity of the resulting energy function has been compared with the specificity of simpler potentials using gapless threading and the ability to predict specific geometry of protein fragments. Significant improvement in threading sensitivity and in the ability to generate sequence-specific protein-like conformations has been achieved.

Microbial genomes have over 72% structure assignment by the threading algorithm PROSPECTOR_Q

The genome scale threading of five complete microbial genomes is revisited using our state-of-the-art threading algorithm, PROSPECTOR_Q. Considering that structure assignment to an ORF could be useful for predicting biochemical function as well as for analyzing pathways, it is important to assess the current status of genome scale threading. The fraction of ORFs to which we could assign protein structures with a reasonably good confidence level to each genome sequences is over 72%, which is significantly higher than earlier studies. Using the assigned structures, we have predicted the function of several ORFs through "single-function" template structures, obtained from an analysis of the relationship between protein fold and function. The fold distribution of the genomes and the effect of the number of homologous sequences on structure assignment are also discussed.

Unfolding of globular proteins: monte carlo dynamics of a realistic reduced model

Reduced lattice models of proteins and Monte Carlo dynamics were used to simulate the initial stages of the unfolding of several proteins of various structural types, and the results were compared to experiment. The models semiquantitatively reproduce the approximate order of events of unfolding as well as subtle mutation effects and effects resulting from differences in sequences of similar folds. The short-time mobility of particular residues, observed in simulations, correlates with the crystallographic temperature factor. The main factor controlling unfolding is the native state topology, with sequence playing a less important role. The correlation with various experiments, especially for sequence-specific effects, strongly suggests that properly designed reduced models of proteins can be used for qualitative studies (or prediction) of protein unfolding pathways.

DPANN: Improved sequence to structure alignments following fold recognition

In fold recognition (FR) a protein sequence of unknown structure is assigned to the closest known three-dimensional (3D) fold. Although FR programs can often identify among all possible folds the one a sequence adopts, they frequently fail to align the sequence to the equivalent residue positions in that fold. Such failures frustrate the next step in structure prediction, protein model building. Hence it is desirable to improve the quality of the alignments between the sequence and the identified structure. We have used artificial neural networks (ANN) to derive a substitution matrix to create alignments between a protein sequence and a protein structure through dynamic programming (DPANN: Dynamic Programming meets Artificial Neural Networks). The matrix is based on the amino acid type and the secondary structure state of each residue. In a database of protein pairs that have the same fold but lack sequences-similarity, DPANN aligns over 30% of all sequences to the paired structure, resembling closely the structural superposition of the pair. In over half of these cases the DPANN alignment is close to the structural superposition, although the initial alignment from the step of fold recognition is not close. Conversely, the alignment created during fold recognition outperforms DPANN in only 10% of all cases. Thus application of DPANN after fold recognition leads to substantial improvements in alignment accuracy, which in turn provides more useful templates for the modeling of protein structures. In the artificial case of using actual instead of predicted secondary structures for the probe protein, over 50% of the alignments are successful.

TOUCHSTONEX: protein structure prediction with sparse NMR data

TOUCHSTONEX, a new method for folding proteins that uses a small number of long-range contact restraints derived from NMR experimental NOE (nuclear Overhauser enhancement) data, is described. The method employs a new lattice-based, reduced model of proteins that explicitly represents C(alpha), C(beta), and the sidechain centers of mass. The force field consists of knowledge-based terms to produce protein-like behavior, including various short-range interactions, hydrogen bonding, and one-body, pairwise, and multibody long-range interactions. Contact restraints were incorporated into the force field as an NOE-specific pairwise potential. We evaluated the algorithm using a set of 125 proteins of various secondary structure types and lengths up to 174 residues. Using N/8 simulated, long-range sidechain contact restraints, where N is the number of residues, 108 proteins were folded to a C(alpha)-root-mean-square deviation (RMSD) from native below 6.5 A. The average RMSD of the lowest RMSD structures for all 125 proteins (folded and unfolded) was 4.4 A. The algorithm was also applied to limited experimental NOE data generated for three proteins. Using very few experimental sidechain contact restraints, and a small number of sidechain-main chain and main chain-main chain contact restraints, we folded all three proteins to low-to-medium resolution structures. The algorithm can be applied to the NMR structure determination process or other experimental methods that can provide tertiary restraint information, especially in the early stage of structure determination, when only limited data are available.

Addition of side chains to a known backbone with defined side-chain centroids

An automatic procedure is proposed for adding side chains to a protein backbone; it is based on optimization of a simplified energy function for peptide side chains, given its backbone and positions of side-chain centroids. The energy is expressed as a sum of the energies of interaction between side chains, and a harmonic penalty function accounting for the preservation of the positions of the C(alpha) atoms and the side-chain centroids. The energy of side-chain interactions is calculated with the soft-sphere ECEPP/3 potential. A Monte Carlo search is carried out to explore all possible side-chain orientations within a fixed backbone and side-chain centroid positions. The initial, usually extended, side-chain conformations are taken directly from the ECEPP/3 database. The procedure was tested on six experimental (X-ray or NMR) structures: immunoglobulin binding protein (PDB code 1IGD, an alpha+beta-protein); transcription factor PML (PDB code 1BOR, a 49-104 fragment of the ring finger domain, predominantly beta-protein); bovine pancreatic trypsin inhibitor (crystal form II) (PDB code 1BPI, an alpha+beta-protein); the monomer of human deoxyhemoglobin (PDB code 1BZ0, an alpha-helical structure); chain A of alcohol dehydrogenase from Drosophila lebanonensis (PDB code 1A4U); as well as on the 10-55 portion of the B domain of staphylococcal protein A (PDB code 1BDD). In all cases except 1BPI, the data for the algorithm (i.e. the backbone or C(alpha) coordinates and the positions of side-chain centroids) were taken from the experimental structures. For protein A, the C(alpha) coordinates and positions of side-chain centroids were also taken from the 1.9-A-resolution model predicted by the UNRES force field. In all comparisons with experimental structures, complete side-chain geometry was reconstructed with a root-mean-square (RMS) deviation of approximately 0.6-0.9 A from the heavy atoms when complete backbone and side-chain-centroid coordinates were used in reconstruction, or approximately 1.0 A when the C(alpha) and centroid coordinates were used.

On the Role of Structural Information in Remote Homology Detection and Sequence Alignment: New Methods Using Hybrid Sequence Profiles

Structural alignments often reveal relationships between proteins that cannot be detected using sequence alignment alone. However, profile search methods based entirely on structural alignments alone have not been found to be effective in finding remote homologs. Here, we explore the role of structural information in remote homolog detection and sequence alignment. To this end, we develop a series of hybrid multidimensional alignment profiles that combine sequence, secondary and tertiary structure information into hybrid profiles. Sequence-based profiles are profiles whose position-specific scoring matrix is derived from sequence alignment alone; structure-based profiles are those derived from multiple structure alignments. We compare pure sequence-based profiles to pure structure-based profiles, as well as to hybrid profiles that use combined sequence-and-structure-based profiles, where sequence-based profiles are used in loop/motif regions and structural information is used in core structural regions. All of the hybrid methods offer significant improvement over simple profile-to-profile alignment. We demonstrate that both sequence-based and structure-based profiles contribute to remote homology detection and alignment accuracy, and that each contains some unique information. We discuss the implications of these results for further improvements in amino acid sequence and structural analysis.

Use of residual dipolar couplings as restraints inab initio protein structure prediction

NMR residual dipolar couplings (RDCs), in the form of the projection angles between the respective internuclear bond vectors, are used as structural restraints in the ab initio structure prediction of a test set of six proteins. The restraints are applied using a recently developed SICHO (SIde-CHain-Only) lattice protein model that employs a replica exchange Monte Carlo (MC) algorithm to search conformational space. Using a small number of RDC restraints, the quality of the predicted structures is improved as reflected by lower RMSD/dRMSD (root mean square deviation/distance root mean square deviation) values from the corresponding native structures and by the higher correlation of the most cooperative mode of motion of each predicted structure with that of the native structure. The latter, in particular, has possible implications for the structure-based functional analysis of predicted structures.

TOUCHSTONE: A unified approach to protein structure prediction

We have applied the TOUCHSTONE structure prediction algorithm that spans the range from homology modeling to ab initio folding to all protein targets in CASP5. Using our threading algorithm PROSPECTOR that does not utilize input from metaservers, one threads against a representative set of PDB templates. If a template is significantly hit, Generalized Comparative Modeling designed to span the range from closely to distantly related proteins from the template is done. This involves freezing the aligned regions and relaxing the remaining structure to accommodate insertions or deletions with respect to the template. For all targets, consensus predicted side chain contacts from at least weakly threading templates are pooled and incorporated into ab initio folding. Often, TOUCHSTONE performs well in the CM to FR categories, with PROSPECTOR showing significant ability to identify analogous templates. When ab initio folding is done, frequently the best models are closer to the native state than the initial template. Among the particularly good predictions are T0130 in the CM/FR category, T0138 in the FR(H) category, T0135 in the FR(A) category, T0170 in the FR/NF category and T0181 in the NF category. Improvements in the approach are needed in the FR/NF and NF categories. Nevertheless, TOUCHSTONE was one of the best performing algorithms over all categories in CASP5.

Structural genomics: computational methods for structure analysis

The success of structural genomics initiatives requires the development and application of tools for structure analysis, prediction, and annotation. In this paper we review recent developments in these areas; specifically structure alignment, the detection of remote homologs and analogs, homology modeling and the use of structures to predict function. We also discuss various rationales for structural genomics initiatives. These include the structure-based clustering of sequence space and genome-wide function assignment. It is also argued that structural genomics can be integrated into more traditional biological research if specific biological questions are included in target selection strategies.

TOUCHSTONE II: a new approach to ab initio protein structure prediction

We have developed a new combined approach for ab initio protein structure prediction. The protein conformation is described as a lattice chain connecting C(alpha) atoms, with attached C(beta) atoms and side-chain centers of mass. The model force field includes various short-range and long-range knowledge-based potentials derived from a statistical analysis of the regularities of protein structures. The combination of these energy terms is optimized through the maximization of correlation for 30 x 60,000 decoys between the root mean square deviation (RMSD) to native and energies, as well as the energy gap between native and the decoy ensemble. To accelerate the conformational search, a newly developed parallel hyperbolic sampling algorithm with a composite movement set is used in the Monte Carlo simulation processes. We exploit this strategy to successfully fold 41/100 small proteins (36 approximately 120 residues) with predicted structures having a RMSD from native below 6.5 A in the top five cluster centroids. To fold larger-size proteins as well as to improve the folding yield of small proteins, we incorporate into the basic force field side-chain contact predictions from our threading program PROSPECTOR where homologous proteins were excluded from the data base. With these threading-based restraints, the program can fold 83/125 test proteins (36 approximately 174 residues) with structures having a RMSD to native below 6.5 A in the top five cluster centroids. This shows the significant improvement of folding by using predicted tertiary restraints, especially when the accuracy of side-chain contact prediction is >20%. For native fold selection, we introduce quantities dependent on the cluster density and the combination of energy and free energy, which show a higher discriminative power to select the native structure than the previously used cluster energy or cluster size, and which can be used in native structure identification in blind simulations. These procedures are readily automated and are being implemented on a genomic scale.

Hierarchical protein folding pathways: a computational study of protein fragments

We have previously presented a building block folding model. The model postulates that protein folding is a hierarchical top-down process. The basic unit from which a fold is constructed, referred to as a hydrophobic folding unit, is the outcome of combinatorial assembly of a set of "building blocks." Results obtained by the computational cutting procedure yield fragments that are in agreement with those obtained experimentally by limited proteolysis. Here we show that as expected, proteins from the same family give very similar building blocks. However, different proteins can also give building blocks that are similar in structure. In such cases the building blocks differ in sequence, stability, contacts with other building blocks, and in their 3D locations in the protein structure. This result, which we have repeatedly observed in many cases, leads us to conclude that while a building block is influenced by its environment, nevertheless, it can be viewed as a stand-alone unit. For small-sized building blocks existing in multiple conformations, interactions with sister building blocks in the protein will increase the population time of the native conformer. With this conclusion in hand, it is possible to develop an algorithm that predicts the building block assignment of a protein sequence whose structure is unknown. Toward this goal, we have created sequentially nonredundant databases of building block sequences. A protein sequence can be aligned against these, in order to be matched to a set of potential building blocks.

Molecular dynamics simulations of peptides and proteins with a continuum electrostatic model based on screened Coulomb potentials

A continuum electrostatics approach for molecular dynamics (MD) simulations of macromolecules is presented and analyzed for its performance on a peptide and a globular protein. The approach incorporates the screened Coulomb potential (SCP) continuum model of electrostatics, which was reported earlier. The model was validated in a broad set of tests some of which were based on Monte Carlo simulations that included single amino acids, peptides, and proteins. The implementation for large-scale MD simulations presented in this article is based on a pairwise potential that makes the electrostatic model suitable for fast analytical calculation of forces. To assess the suitability of the approach, a preliminary validation is conducted, which consists of (i) a 3-ns MD simulation of the immunoglobulin-binding domain of streptococcal protein G, a 56-residue globular protein and (ii) a 3-ns simulation of Dynorphin, a biological peptide of 17 amino acids. In both cases, the results are compared with those obtained from MD simulations using explicit water (EW) molecules in an all-atom representation. The initial structure of Dynorphin was assumed to be an alpha-helix between residues 1 and 9 as suggested from NMR measurements in micelles. The results obtained in the MD simulations show that the helical structure collapses early in the simulation, a behavior observed in the EW simulation and consistent with spectroscopic data that suggest that the peptide may adopt mainly an extended conformation in water. The dynamics of protein G calculated with the SCP implicit solvent model (SCP-ISM) reveals a stable structure that conserves all the elements of secondary structure throughout the entire simulation time. The average structures calculated from the trajectories with the implicit and explicit solvent models had a cRMSD of 1.1 A, whereas each average structure had a cRMSD of about 0.8A with respect to the X-ray structure. The main conformational differences of the average structures with respect to the crystal structure occur in the loop involving residues 8-14. Despite the overall similarity of the simulated dynamics with EW and SCP models, fluctuations of side-chains are larger when the implicit solvent is used, especially in solvent exposed side-chains. The MD simulation of Dynorphin was extended to 40 ns to study its behavior in an aqueous environment. This long simulation showed that the peptide has a tendency to form an alpha-helical structure in water, but the stabilization free energy is too weak, resulting in frequent interconversions between random and helical conformations during the simulation time. The results reported here suggest that the SCP implicit solvent model is adequate to describe electrostatic effects in MD simulation of both peptides and proteins using the same set of parameters. It is suggested that the present approach could form the basis for the development of a reliable and general continuum approach for use in molecular biology, and directions are outlined for attaining this long-term goal.

Fine-grained protein fold assignment by support vector machines using generalized npeptide coding schemes and jury voting from multiple-parameter sets

In the coarse-grained fold assignment of major protein classes, such as all-alpha, all-beta, alpha + beta, alpha/beta proteins, one can easily achieve high prediction accuracy from primary amino acid sequences. However, the fine-grained assignment of folds, such as those defined in the Structural Classification of Proteins (SCOP) database, presents a challenge due to the larger amount of folds available. Recent study yielded reasonable prediction accuracy of 56.0% on an independent set of 27 most populated folds. In this communication, we apply the support vector machine (SVM) method, using a combination of protein descriptors based on the properties derived from the composition of n-peptide and jury voting, to the fine-grained fold prediction, and are able to achieve an overall prediction accuracy of 69.6% on the same independent set-significantly higher than the previous results. On 10-fold cross-validation, we obtained a prediction accuracy of 65.3%. Our results show that SVM coupled with suitable global sequence-coding schemes can significantly improve the fine-grained fold prediction. Our approach should be useful in structure prediction and modeling.

Ab initio protein structure prediction via a combination of threading, lattice folding, clustering, and structure refinement

A combination of sequence comparison, threading, lattice, and off-lattice Monte Carlo (MC) simulations and clustering of MC trajectories was used to predict the structure of all (but one) targets of the CASP4 experiment on protein structure prediction. Although this method is automated and is operationally the same regardless of the level of uniqueness of the query proteins, here we focus on the more difficult targets at the border of the fold recognition and new fold categories. For a few targets (T0110 is probably the best example), the ab initio method produced more accurate models than models obtained by the fold recognition techniques. For the most difficult targets from the new fold categories, substantial fragments of structures have been correctly predicted. Possible improvements of the method are briefly discussed.