Comparative modelling by restraint-based conformational sampling

Background  

Although comparative modelling is routinely used to produce three-dimensional models of proteins, very few automated approaches are formulated in a way that allows inclusion of restraints derived from experimental data as well as those from the structures of homologues. Furthermore, proteins are usually described as a single conformer, rather than an ensemble that represents the heterogeneity and inaccuracy of experimentally determined protein structures. Here we address these issues by exploring the application of the restraint-based conformational space search engine, RAPPER, which has previously been developed for rebuilding experimentally defined protein structures and for fitting models to electron density derived from X-ray diffraction analyses.     

Solid-state NMR structures of integral membrane proteins

Abstract

Solid-state NMR is unique for its ability to obtain three-dimensional structures and to measure atomic-resolution structural and dynamic information for membrane proteins in native lipid bilayers. An increasing number and complexity of integral membrane protein structures have been determined by solid-state NMR using two main methods. Oriented sample solid-state NMR uses macroscopically aligned lipid bilayers to obtain orientational restraints that define secondary structure and global fold of embedded peptides and proteins and their orientation and topology in lipid bilayers. Magic angle spinning (MAS) solid-state NMR uses unoriented rapidly spinning samples to obtain distance and torsion angle restraints that define tertiary structure and helix packing arrangements. Details of all current protein structures are described, highlighting developments in experimental strategy and other technological advancements. Some structures originate from combining solid- and solution-state NMR information and some have used solid-state NMR to refine X-ray crystal structures. Solid-state NMR has also validated the structures of proteins determined in different membrane mimetics by solution-state NMR and X-ray crystallography and is therefore complementary to other structural biology techniques. By continuing efforts in identifying membrane protein targets and developing expression, isotope labelling and sample preparation strategies, probe technology, NMR experiments, calculation and modelling methods and combination with other techniques, it should be feasible to determine the structures of many more membrane proteins of biological and biomedical importance using solid-state NMR. This will provide three-dimensional structures and atomic-resolution structural information for characterising ligand and drug interactions, dynamics and molecular mechanisms of membrane proteins under physiological lipid bilayer conditions.     

Correlation of 4Pi and Electron Microscopy to Study Transport Through Single Golgi Stacks in Living Cells with Super Resolution

Two problems have hampered the use of light microscopy for structural studies of cellular organelles for a long time: the limited resolution and the difficulty of obtaining true structural boundaries from complex intensity curves. The advent of modern high-resolution light microscopy techniques and their combination with objective image segmentation now provide us with the means to bridge the gap between light and electron microscopy in cell biology applications. In this study, we provide the first comparative correlative analysis of three-dimensional structures obtained by 4Pi microscopy and segmented by a zero-crossing procedure with those of transmission electron microscopy (TEM). The distribution within the cisternae of isolated Golgi stacks of the cargo protein procollagen 3 was mapped by both 4Pi microscopy and TEM for a detailed comparative analysis of their imaging capabilities. A high correlation was seen for the structures, indicating the particular accuracy of the 4Pi microscopy. Furthermore, for the first time, transport of a cargo molecule (vesicular stomatitis virus G protein-pEGFP) through individual Golgi stacks (labeled by galactosyl transferase-venusYFP) was visualized by 4Pi microscopy. Following the procedures validated by the correlative analysis, our transport experiments show that (i) VSVG-pEGFP rapidly enter/exit individual Golgi stacks, (ii) VSVG-pEGFP never fills the GalT-venusYFP compartments completely and (iii) the GalT-venusYFP compartment volume increases upon VSVG-pEGFP arrival. This morphological evidence supports some previous TEM-based observations of intra-Golgi transport of VSVG-pEGFP and provides new insights toward a better understanding of protein progression across Golgi stacks. Our study thus demonstrates the general applicability of super-resolution fluorescence microscopy, coupled with the zero-crossing segmentation procedure, for structural studies of suborganelle protein distributions under living cell conditions.     

The Genetic Algorithm Applied as a Modelling Tool to Predict the Fold of Small Proteins with Different Topologies

The genetic algorithm exploits the principles of natural evolution. Solution trials are evolved by mutation, recombination and selection until they achieve near optimal solutions [1].Our own approach has now been developed [2] after a general overview on the application potential for protein structure analysis [3] to a tool to delineate the three-dimensional topology for the mainchain of small proteins [4], no matter whether they are largely helical, are mixed or -strand rich [5].Results on several protein examples for these different modelling tasks are presented and compared with the experimentally observed structures (RMSDs are around 4.5-5.5 Å). To start a modelling trial only the protein sequence and knowledge of its secondary structure is required. The fittest folds obtained after the evolution at the end of the simulations yield the three dimensional models of the fold. Current limitations are protein size (generally less than 100 aminoacids), number of secondary structure elements [7-8] and irregular topologies (e.g. ferridoxins).Further, preliminary results from current simulations are illustrated. We now want to apply simple experimental or other information, which is available long before the three-dimensional structure of the protein becomes known, to refine the modelling of the protein fold and tackle also more difficult modelling examples by our tool.Supplementary material to this paper is available in electronic form at http://dx.doi.org/10.1007/s0089460020304     

Outlining folding nuclei in globular proteins

Our theoretical approach for prediction of folding/unfolding nuclei in three-dimensional protein structures is based on a search for free energy saddle points on networks of protein unfolding pathways. Under some approximations, this search is performed rapidly by dynamic programming and results in prediction of Phi values, which can be compared with those found experimentally. In this study, we optimize some details of the model (specifically, hydrogen atoms are taken into account in addition to heavy atoms), and compare the theoretically obtained and experimental Phi values (which characterize involvement of residues in folding nuclei) for all 17 proteins, where Phi values are now known for many residues. We show that the model provides good Phi value predictions for proteins whose structures have been determined by X-ray analysis (the average correlation coefficient is 0.65), with a more limited success for proteins whose structures have been determined by NMR techniques only (the average correlation coefficient is 0.34), and that the transition state free energies computed from the same model are in a good anticorrelation with logarithms of experimentally measured folding rates at mid-transition (the correlation coefficient is -0.73).     

Bridging from molecular simulation to biochemical networks

How can we make the connection between the three-dimensional structures of individual proteins and understanding how complex biological systems involving many proteins work? The modelling and simulation of protein structures can help to answer this question for systems ranging from multimacromolecular complexes to organelles and cells. On one hand, multiscale modelling and simulation techniques are advancing to permit the spatial and temporal properties of large systems to be simulated using atomic-detail structures. On the other hand, the estimation of kinetic parameters for the mathematical modelling of biochemical pathways using protein structure information provides a basis for iterative manipulation of biochemical pathways guided by protein structure. Recent advances include the structural modelling of protein complexes on the genomic level, novel coarse-graining strategies to increase the size of the system and the time span that can be simulated, and comparative molecular field analyses to estimate enzyme kinetic parameters.     

Apolipoprotein structural organization in high density lipoproteins: belts, bundles, hinges and hairpins

PURPOSE OF REVIEW: To summarize recent advances towards an understanding of the three-dimensional structures of the apolipoprotein components of HDL with a specific focus on high resolution models of apolipoprotein A-I. RECENT FINDINGS: Since the primary sequence was first reported, various models have been advanced for the structure of apolipoprotein A-I, the major protein constituent of HDL, in its lipid-free and lipid-bound forms. Unfortunately, the generation of experimental data capable of distinguishing among the competing models has lagged far behind. However, recent experimental strategies, including X-ray crystallography, applications of resonance energy transfer and mass spectrometry, have combined with sophisticated theoretical approaches to develop three-dimensional structural models of apolipoprotein A-I with previously unavailable resolution. SUMMARY: The recent synergy of sophisticated computer modeling techniques with hard experimental data has generated new models for apolipoprotein A-I in certain subclasses of HDL produced in vitro. The challenge now is to adapt and test these models in the more complex forms of HDL isolated directly from human plasma.     

Online homology modelling as a means of bridging the sequence-structure gap

For even the best-studied species, there is a large gap in their representation in the protein databank (PDB) compared to within sequence databases. Typically, less than 2% of sequences are represented in the PDB. This is partly due to the considerable experimental challenge and manual inputs required to solve three dimensional structures by methods such as X-ray diffraction and multi-dimensional nuclear magnetic resonance (NMR) spectroscopy in comparison to high-throughput sequencing. This gap is made even wider by the high level of redundancy within the PDB and under-representation of some protein categories such as membrane-associated proteins which comprise approximately 25% of proteins encoded in genomes. A traditional route to closing the sequence-structure gap is offered by homology modelling whereby the sequence of a target protein is modelled on a template represented in the PDB using in silico energy minimisation approaches. More recently, online homology servers have become available which automatically generate models from proffered sequences. However, many online servers give little indication of the structural plausibility of the generated model. In this paper, the online homology server Geno3D will be described. This server uses similar software to that used in modelling structures during structure determination and thus generates data allowing determination of the structural plausibility of models. For illustration, modelling of a chemotaxis protein (CheY) from Pseudomononas entomophila L48 (accession YP_609298) on a template (PDB id. 1mvo), the phosphorylation domain of an outer membrane protein PhoP from Bacillus subtilis, will be described.     

Prediction of interface residues in protein-protein complexes by a consensus neural network method: test against NMR data

The number of structures of protein-protein complexes deposited to the Protein Data Bank is growing rapidly. These structures embed important information for predicting structures of new protein complexes. This motivated us to develop the PPISP method for predicting interface residues in protein-protein complexes. In PPISP, sequence profiles and solvent accessibility of spatially neighboring surface residues were used as input to a neural network. The network was trained on native interface residues collected from the Protein Data Bank. The prediction accuracy at the time was 70% with 47% coverage of native interface residues. Now we have extensively improved PPISP. The training set now consisted of 1156 nonhomologous protein chains. Test on a set of 100 nonhomologous protein chains showed that the prediction accuracy is now increased to 80% with 51% coverage. To solve the problem of over-prediction and under-prediction associated with individual neural network models, we developed a consensus method that combines predictions from multiple models with different levels of accuracy and coverage. Applied on a benchmark set of 68 proteins for protein-protein docking, the consensus approach outperformed the best individual models by 3-8 percentage points in accuracy. To demonstrate the predictive power of cons-PPISP, eight complex-forming proteins with interfaces characterized by NMR were tested. These proteins are nonhomologous to the training set and have a total of 144 interface residues identified by chemical shift perturbation. cons-PPISP predicted 174 interface residues with 69% accuracy and 47% coverage and promises to complement experimental techniques in characterizing protein-protein interfaces. .     

Recognition of errors in three-dimensional structures of proteins

A major problem in the determination of the three-dimensional structure of proteins concerns the quality of the structural models obtained from the interpretation of experimental data. New developments in X-ray crystallography and nuclear magnetic resonance spectroscopy have acceleratedd the process of structure determination and the biological community is confronted with a steadily increasing number of experimentally determined protein folds. However, in the recent past several experimentally determined protein structures have been proven to contain major errors, indicating that in some cases the interpretation of experimental data is difficult and may yield incorrect models. Such problems can be avoided when computational methods are employed which complement experimental structure determinations. A prerequisite of such computational tools is that they are independent of the parameters obtained from a particular experiment. In addition such techniques are able to support and accelerate experimental structure determinations. Here we present techniques based on knowledge based mean fields which can be used to judge the quality of protein folds. The methods can be used to identify misfolded structures as well as faulty parts of structural models. The techniques are even applicable in cases where only the C_α trace of a protein conformation is available. The capabilities of the technique are demonstrated using correct and incorrect protein folds. © 1993 Wiley-Liss, Inc.     

The SWISS-MODEL workspace: a web-based environment for protein structure homology modelling

MOTIVATION: Homology models of proteins are of great interest for planning and analysing biological experiments when no experimental three-dimensional structures are available. Building homology models requires specialized programs and up-to-date sequence and structural databases. Integrating all required tools, programs and databases into a single web-based workspace facilitates access to homology modelling from a computer with web connection without the need of downloading and installing large program packages and databases. RESULTS: SWISS-MODEL workspace is a web-based integrated service dedicated to protein structure homology modelling. It assists and guides the user in building protein homology models at different levels of complexity. A personal working environment is provided for each user where several modelling projects can be carried out in parallel. Protein sequence and structure databases necessary for modelling are accessible from the workspace and are updated in regular intervals. Tools for template selection, model building and structure quality evaluation can be invoked from within the workspace. Workflow and usage of the workspace are illustrated by modelling human Cyclin A1 and human Transmembrane Protease 3. AVAILABILITY: The SWISS-MODEL workspace can be accessed freely at http://swissmodel.expasy.org/workspace/     

An improved protein structure evaluation using a semi-empirically derived structure property

Background

In the backdrop of challenge to obtain a protein structure under the known limitations of both experimental and theoretical techniques, the need of a fast as well as accurate protein structure evaluation method still exists to substantially reduce a huge gap between number of known sequences and structures. Among currently practiced theoretical techniques, homology modelling backed by molecular dynamics based optimization appears to be the most popular one. However it suffers from contradictory indications of different validation parameters generated from a set of protein models which are predicted against a particular target protein. For example, in one model Ramachandran Score may be quite high making it acceptable, whereas, its potential energy may not be very low making it unacceptable and vice versa. Towards resolving this problem, the main objective of this study was fixed as to utilize a simple experimentally derived output, Surface Roughness Index of concerned protein of unknown structure as an intervening agent that could be obtained using ordinary microscopic images of heat denatured aggregates of the same protein.

Result

It was intriguing to observe that direct experimental knowledge of the concerned protein, however simple it may be, might give insight on acceptability of its particular structural model out of a confusion set of models generated from database driven comparative technique for structure prediction. The result obtained from a widely varying structural class of proteins indicated that speed of protein structure evaluation can be further enhanced without compromising with accuracy by recruiting simple experimental output.

Conclusion

In this work, a semi-empirical methodological approach was provided for improving protein structure evaluation. It showed that, once structure models of a protein were obtained through homology technique, the problem of selection of a best model out of a confusion set of Pareto-optimal structures could be resolved by employing a structure agent directly obtainable through experiment with the same protein as experimental ingredient. Overall, in the backdrop of getting a reasonably accurate protein structure of pathogens causing epidemics or biological warfare, such approach could be of use as a plausible solution for fast drug design.

     

About the use of protein models

Protein models can be of great assistance in functional genomics, as they provide the structural insights often necessary to understand protein function. Although comparative modelling is far from yielding perfect structures, this is still the most reliable method and the quality of the predictions is now well understood. Models can be classified according to their correctness and accuracy, which will impact their applicability and usefulness in functional genomics and a variety of situations.     

Revisiting gap locations in amino acid sequence alignments and a proposal for a method to improve them by introducing solvent accessibility

In comparative modeling, the quality of amino acid sequence alignment still constitutes a major bottleneck in the generation of high quality models of protein three-dimensional (3D) structures. Substantial efforts have been made to improve alignment quality by revising the substitution matrix, introducing multiple sequences, replacing dynamic programming with hidden Markov models, and incorporating 3D structure information. Improvements in the gap penalty have not been a major focus, however, following the development of the affine gap penalty and of the secondary structure dependent gap penalty. We revisited the correlation between protein 3D structure and gap location in a large protein 3D structure data set, and found that the frequency of gap locations approximated to an exponential function of the solvent accessibility of the inserted residues. The nonlinearity of the gap frequency as a function of accessibility corresponded well to the relationship between residue mutation pattern and residue accessibility. By introducing this relationship into the gap penalty calculation for pairwise alignment between template and target amino acid sequences, we were able to obtain a sequence alignment much closer to the structural alignment. The quality of the alignments was substantially improved on a pair of sequences with identity in the "twilight zone" between 20 and 40%. The relocation of gaps by our new method made a significant improvement in comparative modeling, exemplified here by the Bacillus subtilis yitF protein. The method was implemented in a computer program, ALAdeGAP (ALignment with Accessibility dependent GAp Penalty), which is available at http://cib.cf.ocha.ac.jp/target_protein/.     

Progress in protein structure prediction: assessment of CASP3.

The third comparative assessment of techniques of protein structure prediction (CASP3) was held during 1998. This is a blind trial in which structures are predicted prior to having knowledge of the coordinates, which are then revealed to enable the assessment. Three sections at the meeting evaluated different methodologies - comparative modelling, fold recognition and ab initio methods. For some, but not all of the target coordinates, high quality models were submitted in each of these sections. There have been improvements in prediction techniques since CASP2 in 1996, most notably for ab initio methods.     

Evaluation of accuracy and applicability of protein models: retrospective analysis of biological and biomedical predictions

In order to study protein function and activity structural data is required. Since experimental structures are available for just a small fraction of all known protein sequences, computational methods such as protein modelling can provide useful information. Over the last few decades we have predicted, with homology modelling methods, the structures for numerous proteins. In this study we assess the structural quality and validity of the biological and medical interpretations and predictions made based on the models. All the models had correct scaffolding and were ranked at least as correct or good by numerical evaluators even though the sequence identity with the template was as low as 8%. The biological explanations made based on models were well in line with experimental structures and other experimental studies. Retrospective analysis of homology models indicates the power of protein modelling when made carefully from sequence alignment to model building and refinement. Modelling can be applied to studying and predicting different kinds of biological phenomena and according to our results it can be done so with success.     

Homology modeling: an important tool for the drug discovery

In the last decades, homology modeling has become a popular tool to access theoretical three-dimensional (3D) structures of molecular targets. So far several 3D models of proteins have been built by this technique and used in a great diversity of structural biology studies. But are those models consistent enough with experimental structures to make this technique an effective and reliable tool for drug discovery? Here we present, briefly, the fundamentals and current state-of-the-art of the homology modeling techniques used to build 3D structures of molecular targets, which experimental structures are not available in databases, and list some of the more important works, using this technique, available in literature today. In many cases those studies have afforded successful models for the drug design of more selective agonists/antagonists to the molecular targets in focus and guided promising experimental works, proving that, when the appropriate templates are available, useful models can be built using some of the several software available today for this purpose. Limitations of the experimental techniques used to solve 3D structures allied to constant improvements in the homology modeling software will maintain the need for theoretical models, establishing the homology modeling as a fundamental tool for the drug discovery.     

Probabilistic cross-link analysis and experiment planning for high-throughput elucidation of protein structure

Emerging high-throughput techniques for the characterization of protein and protein-complex structures yield noisy data with sparse information content, placing a significant burden on computation to properly interpret the experimental data. One such technique uses cross-linking (chemical or by cysteine oxidation) to confirm or select among proposed structural models (e.g., from fold recognition, ab initio prediction, or docking) by testing the consistency between cross-linking data and model geometry. This paper develops a probabilistic framework for analyzing the information content in cross-linking experiments, accounting for anticipated experimental error. This framework supports a mechanism for planning experiments to optimize the information gained. We evaluate potential experiment plans using explicit trade-offs among key properties of practical importance: discriminability, coverage, balance, ambiguity, and cost. We devise a greedy algorithm that considers those properties and, from a large number of combinatorial possibilities, rapidly selects sets of experiments expected to discriminate pairs of models efficiently. In an application to residue-specific chemical cross-linking, we demonstrate the ability of our approach to plan experiments effectively involving combinations of cross-linkers and introduced mutations. We also describe an experiment plan for the bacteriophage lambda Tfa chaperone protein in which we plan dicysteine mutants for discriminating threading models by disulfide formation. Preliminary results from a subset of the planned experiments are consistent and demonstrate the practicality of planning. Our methods provide the experimenter with a valuable tool (available from the authors) for understanding and optimizing cross-linking experiments.     

Can molecular dynamics simulations help in discriminating correct from erroneous protein 3D models?

Background  

Recent approaches for predicting the three-dimensional (3D) structure of proteins such asde novoor fold recognition methods mostly rely on simplified energy potential functions and a reduced representation of the polypeptide chain. These simplifications facilitate the exploration of the protein conformational space but do not permit to capture entirely the subtle relationship that exists between the amino acid sequence and its native structure. It has been proposed that physics-based energy functions together with techniques for sampling the conformational space, e.g., Monte Carlo or molecular dynamics (MD) simulations, are better suited to the task of modelling proteins at higher resolutions than those of models obtained with the former type of methods. In this study we monitor different protein structural properties along MD trajectories to discriminate correct from erroneous models. These models are based on the sequence-structure alignments provided by our fold recognition method, FROST. We define correct models as being built from alignments of sequences with structures similar to their native structures and erroneous models from alignments of sequences with structures unrelated to their native structures.     

Predicting structures for genome proteins.

Assigning three-dimensional protein folds to genome sequences is essential to understanding protein function. Although experimental three-dimensional structures are currently available for only a very small fraction of these sequences, computational fold assignment is able to assign folds to 20-30% of the sequences in various genomes. This percentage varies depending on the particular organism under analysis, on the sensitivities of the methods used and on the number of experimental structures available at the time the assignment is carried out. The fraction of assignable sequences is currently increasing at an annual rate of roughly 18%. If this rate is sustained throughout the coming years, three-dimensional computational models for more than half of the genome sequences may be available by the year 2003.