Three-dimensional modeling of the I-TevI homing endonuclease catalytic domain, a GIY-YIG superfamily member, using NMR restraints and Monte Carlo dynamics.

Using a recent version of the SICHO algorithm for in silico protein folding, we made a blind prediction of the tertiary structure of the N-terminal, independently folded, catalytic domain (CD) of the I-TevI homing endonuclease, a representative of the GIY-YIG superfamily of homing endonucleases. The secondary structure of the I-TevI CD has been determined using NMR spectroscopy, but computational sequence analysis failed to detect any protein of known tertiary structure related to the GIY-YIG nucleases (Kowalski et al., Nucleic Acids Res., 1999, 27, 2115-2125). To provide further insight into the structure-function relationships of all GIY-YIG superfamily members, including the recently described subfamily of type II restriction enzymes (Bujnicki et al., Trends Biochem. Sci., 2000, 26, 9-11), we incorporated the experimentally determined and predicted secondary and tertiary restraints in a reduced (side chain only) protein model, which was minimized by Monte Carlo dynamics and simulated annealing. The subsequently elaborated full atomic model of the I-TevI CD allows the available experimental data to be put into a structural context and suggests that the GIY-YIG domain may dimerize in order to bring together the conserved residues of the active site.

TOUCHSTONE: an ab initio protein structure prediction method that uses threading-based tertiary restraints.

The successful prediction of protein structure from amino acid sequence requires two features: an efficient conformational search algorithm and an energy function with a global minimum in the native state. As a step toward addressing both issues, a threading-based method of secondary and tertiary restraint prediction has been developed and applied to ab initio folding. Such restraints are derived by extracting consensus contacts and local secondary structure from at least weakly scoring structures that, in some cases, can lack any global similarity to the sequence of interest. Furthermore, to generate representative protein structures, a reduced lattice-based protein model is used with replica exchange Monte Carlo to explore conformational space. We report results on the application of this methodology, termed TOUCHSTONE, to 65 proteins whose lengths range from 39 to 146 residues. For 47 (40) proteins, a cluster centroid whose rms deviation from native is below 6.5 (5) A is found in one of the five lowest energy centroids. The number of correctly predicted proteins increases to 50 when atomic detail is added and a knowledge-based atomic potential is combined with clustered and nonclustered structures for candidate selection. The combination of the ratio of the relative number of contacts to the protein length and the number of clusters generated by the folding algorithm is a reliable indicator of the likelihood of successful fold prediction, thereby opening the way for genome-scale ab initio folding.

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

An improved generalized comparative modeling method, GENECOMP, for the refinement of threading models is developed and validated on the Fischer database of 68 probe-template pairs, a standard benchmark used to evaluate threading approaches. The basic idea is to perform ab initio folding using a lattice protein model, SICHO, near the template provided by the new threading algorithm PROSPECTOR. PROSPECTOR also provides predicted contacts and secondary structure for the template-aligned regions, and possibly for the unaligned regions by garnering additional information from other top-scoring threaded structures. Since the lowest-energy structure generated by the simulations is not necessarily the best structure, we employed two structure-selection protocols: distance geometry and clustering. In general, clustering is found to generate somewhat better quality structures in 38 of 68 cases. When applied to the Fischer database, the protocol does no harm and in a significant number of cases improves upon the initial threading model, sometimes dramatically. The procedure is readily automated and can be implemented on a genomic scale.

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.

Derivation of protein-specific pair potentials based on weak sequence fragment similarity.

A method is presented for the derivation of knowledge-based pair potentials that corrects for the various compositions of different proteins. The resulting statistical pair potential is more specific than that derived from previous approaches as assessed by gapless threading results. Additionally, a methodology is presented that interpolates between statistical potentials when no homologous examples to the protein of interest are in the structural database used to derive the potential, to a Go-like potential (in which native interactions are favorable and all nonnative interactions are not) when homologous proteins are present. For cases in which no protein exceeds 30% sequence identity, pairs of weakly homologous interacting fragments are employed to enhance the specificity of the potential. In gapless threading, the mean z score increases from -10.4 for the best statistical pair potential to -12.8 when the local sequence similarity, fragment-based pair potentials are used. Examination of the ab initio structure prediction of four representative globular proteins consistently reveals a qualitative improvement in the yield of structures in the 4 to 6 A rmsd from native range when the fragment-based pair potential is used relative to that when the quasichemical pair potential is employed. This suggests that such protein-specific potentials provide a significant advantage relative to generic quasichemical potentials.

Dynamics and thermodynamics of beta-hairpin assembly: insights from various simulation techniques.

Small peptides that might have some features of globular proteins can provide important insights into the protein folding problem. Two simulation methods, Monte Carlo Dynamics (MCD), based on the Metropolis sampling scheme, and Entropy Sampling Monte Carlo (ESMC), were applied in a study of a high-resolution lattice model of the C-terminal fragment of the B1 domain of protein G. The results provide a detailed description of folding dynamics and thermodynamics and agree with recent experimental findings (. Nature. 390:196-197). In particular, it was found that the folding is cooperative and has features of an all-or-none transition. Hairpin assembly is usually initiated by turn formation; however, hydrophobic collapse, followed by the system rearrangement, was also observed. The denatured state exhibits a substantial amount of fluctuating helical conformations, despite the strong beta-type secondary structure propensities encoded in the sequence.

Ab initio folding of proteins using restraints derived from evolutionary information.

We present our predictions in the ab initio structure prediction category of CASP3. Eleven targets were folded, using a method based on a Monte Carlo search driven by secondary and tertiary restraints derived from multiple sequence alignments. Our results can be qualitatively summarized as follows: The global fold can be considered "correct" for targets 65 and 74, "almost correct" for targets 64, 75, and 77, "half-correct" for target 79, and "wrong" for targets 52, 56, 59, and 63. Target 72 has not yet been solved experimentally. On average, for small helical and alpha/beta proteins (on the order of 110 residues or smaller), the method predicted low resolution structures with a reasonably good prediction of the global topology. Most encouraging is that in some situations, such as with target 75 and, particularly, target 77, the method can predict a substantial portion of a rare or even a novel fold. However, the current method still fails on some beta proteins, proteins over the 110-residue threshold, and sequences in which only a poor multiple sequence alignment can be built. On the other hand, for small proteins, the method gives results of quality at least similar to that of threading, with the advantage of not being restricted to known folds in the protein database. Overall, these results indicate that some progress has been made on the ab initio protein folding problem. Detailed information about our results can be obtained by connecting to http:/(/)www.bioinformatics.danforthcenter.org/+ ++CASP3.

De novo simulations of the folding thermodynamics of the GCN4 leucine zipper.

Entropy Sampling Monte Carlo (ESMC) simulations were carried out to study the thermodynamics of the folding transition in the GCN4 leucine zipper (GCN4-lz) in the context of a reduced model. Using the calculated partition functions for the monomer and dimer, and taking into account the equilibrium between the monomer and dimer, the average helix content of the GCN4-lz was computed over a range of temperatures and chain concentrations. The predicted helix contents for the native and denatured states of GCN4-lz agree with the experimental values. Similar to experimental results, our helix content versus temperature curves show a small linear decline in helix content with an increase in temperature in the native region. This is followed by a sharp transition to the denatured state. van't Hoff analysis of the helix content versus temperature curves indicates that the folding transition can be described using a two-state model. This indicates that knowledge-based potentials can be used to describe the properties of the folded and unfolded states of proteins.

Assembly of protein structure from sparse experimental data: an efficient Monte Carlo model.

A new, efficient method for the assembly of protein tertiary structure from known, loosely encoded secondary structure restraints and sparse information about exact side chain contacts is proposed and evaluated. The method is based on a new, very simple method for the reduced modeling of protein structure and dynamics, where the protein is described as a lattice chain connecting side chain centers of mass rather than Calphas. The model has implicit built-in multibody correlations that simulate short- and long-range packing preferences, hydrogen bonding cooperativity and a mean force potential describing hydrophobic interactions. Due to the simplicity of the protein representation and definition of the model force field, the Monte Carlo algorithm is at least an order of magnitude faster than previously published Monte Carlo algorithms for structure assembly. In contrast to existing algorithms, the new method requires a smaller number of tertiary restraints for successful fold assembly; on average, one for every seven residues as compared to one for every four residues. For example, for smaller proteins such as the B domain of protein G, the resulting structures have a coordinate root mean square deviation (cRMSD), which is about 3 A from the experimental structure; for myoglobin, structures whose backbone cRMSD is 4.3 A are produced, and for a 247-residue TIM barrel, the cRMSD of the resulting folds is about 6 A. As would be expected, increasing the number of tertiary restraints improves the accuracy of the assembled structures. The reliability and robustness of the new method should enable its routine application in model building protocols based on various (very sparse) experimentally derived structural restraints.

Combined multiple sequence reduced protein model approach to predict the tertiary structure of small proteins.

By incorporating predicted secondary and tertiary restraints into ab initio folding simulations, low resolution tertiary structures of a test set of 20 nonhomologous proteins have been predicted. These proteins, which represent all secondary structural classes, contain from 37 to 100 residues. Secondary structural restraints are provided by the PHD secondary structure prediction algorithm that incorporates multiple sequence information. Predicted tertiary restraints are obtained from multiple sequence alignments via a two-step process: First, "seed" side chain contacts are identified from a correlated mutation analysis, and then, the seed contacts are "expanded" by an inverse folding algorithm. These predicted restraints are then incorporated into a lattice based, reduced protein model. Depending upon fold complexity, the resulting nativelike topologies exhibit a coordinate root-mean-square deviation, cRMSD, from native between 3.1 and 6.7 A. Overall, this study suggests that the use of restraints derived from multiple sequence alignments combined with a fold assembly algorithm is a promising approach to the prediction of the global topology of small proteins.

Computer simulations of de novo designed helical proteins.

In the context of reduced protein models, Monte Carlo simulations of three de novo designed helical proteins (four-member helical bundle) were performed. At low temperatures, for all proteins under consideration, protein-like folds having different topologies were obtained from random starting conformations. These simulations are consistent with experimental evidence indicating that these de novo designed proteins have the features of a molten globule state. The results of Monte Carlo simulations suggest that these molecules adopt four-helix bundle topologies. They also give insight into the possible mechanism of folding and association, which occurs in these simulations by on-site assembly of the helices. The low-temperature conformations of all three sequences have the features of a molten globule state.

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

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

Nativelike topology assembly of small proteins using predicted restraints in Monte Carlo folding simulations.

By incorporating predicted secondary and tertiary restraints derived from multiple sequence alignments into ab initio folding simulations, it has been possible to assemble native-like tertiary structures for a test set of 19 nonhomologous proteins ranging from 29 to 100 residues in length and representing all secondary structural classes. Secondary structural restraints are provided by the PHD secondary structure prediction algorithm that incorporates multiple sequence information. Multiple sequence alignments also provide predicted tertiary restraints via a two-step process: First, seed side chain contacts are selected from a correlated mutation analysis, and then an inverse folding algorithm expands these seed contacts. The predicted secondary and tertiary restraints are incorporated into a lattice-based, reduced protein model for structure assembly and refinement. The resulting native-like topologies exhibit a coordinate root-mean-square deviation from native for the whole chain between 3.1 and 6.7 A, with values ranging from 2.6 to 4.1 A over approximately 80% of the structure. Overall, this study suggests that the use of restraints derived from multiple sequence alignments combined with a fold assembly algorithm is a promising approach to the prediction of the global topology of small proteins.

Tertiary structure prediction of the KIX domain of CBP using Monte Carlo simulations driven by restraints derived from multiple sequence alignments.

Using a recently developed protein folding algorithm, a prediction of the tertiary structure of the KIX domain of the CREB binding protein is described. The method incorporates predicted secondary and tertiary restraints derived from multiple sequence alignments in a reduced protein model whose conformational space is explored by Monte Carlo dynamics. Secondary structure restraints are provided by the PHD secondary structure prediction algorithm that was modified for the presence of predicted U-turns, i.e., regions where the chain reverses global direction. Tertiary restraints are obtained via a two-step process: First, seed side-chain contacts are identified from a correlated mutation analysis, and then, a threading-based algorithm expands the number of these seed contacts. Blind predictions indicate that the KIX domain is a putative three-helix bundle, although the chirality of the bundle could not be uniquely determined. The expected root-mean-square deviation for the correct chirality of the KIX domain is between 5.0 and 6.2 A. This is to be compared with the estimate of 12.9 A that would be expected by a random prediction, using the model of F. Cohen and M. Sternberg (J. Mol. Biol. 138:321-333, 1980).

Method for low resolution prediction of small protein tertiary structure.

A new method for the de novo prediction of protein structures at low resolution has been developed. Starting from a multiple sequence alignment, protein secondary structure is predicted, and only those topological elements with high reliability are selected. Then, the multiple sequence alignment and the secondary structure prediction are combined to predict side chain contacts. Such contact map prediction is carried out in two stages. First, an analysis of correlated mutations is carried out to identify pairs of topological elements of secondary structure which are in contact. Then, inverse folding is used to select compatible fragments in contact, thereby enriching the number and identity of predicted side chain contacts. The final outcome of the procedure is a set of noisy secondary and tertiary restraints. These are used as a restrained potential in a Monte Carlo simulation of simplified protein models driven by statistical potentials. Low energy structures are then searched for by using simulated annealing techniques. Implementation of the restraints is carried out so as to take into account of their low resolution. Using this procedure, it has been possible to predict de novo the structure of three very different protein topologies: an alpha/beta protein, the bovine pancreatic trypsin inhibitor (6pti), an alpha-helical protein, calbindin (3icb), and an all beta- protein, the SH3 domain of spectrin (1shg). In all cases, low resolution folds have been obtained with a root mean square deviation (RMSD) of 4.5-5.5 A with respect to the native structure. Some misfolded topologies appear in the simulations, but it is possible to select the native one on energetic grounds. Thus, it is demonstrated that the methodology is general for all protein motifs. Work is in progress in order to test the methodology on a larger set of protein structures.

Folding simulations and computer redesign of protein A three-helix bundle motifs.

In solution, the B domain of protein A from Staphylococcus aureus (B domain) possesses a three-helix bundle structure. This simple motif has been previously reproduced by Kolinski and Skolnick (Proteins 18: 353-366, 1994) using a reduced representation lattice model of proteins with a statistical interaction scheme. In this paper, an improved version of the potential has been used, and the robustness of this result has been tested by folding from the random state a set of three-helix bundle proteins that are highly homologous to the B domain of protein A. Furthermore, an attempt to redesign the B domain native structure to its topological mirror image fold has been made by multiple mutations of the hydrophobic core and the turn region between helices I and II. A sieve method for scanning a large set of mutations to search for this desired property has been proposed. It has been shown that mutations of native B domain hydrophobic core do not introduce significant changes in the protein motif. Mutations in the turn region were also very conservative; nevertheless, a few mutants acquired the desired topological mirror image motif. A set of all atom models of the most probable mutant was reconstructed from the reduced models and refined using a molecular dynamics algorithm in the presence of water. The packing of all atom structures obtained corroborates the lattice model results. We conclude that the change in the handedness of the turn induced by the mutations, augmented by the repacking of hydrophobic core and the additional burial of the second helix N-cap side chain, are responsible for the predicted preferential adoption of the mirror image structure.

An algorithm for prediction of structural elements in small proteins.

A method for predicting the location of surface loops/turns and assigning the intervening secondary structure of the transglobular linkers in small, single domain globular proteins has been developed. Application to a set of 10 proteins of known structure indicates a high level of accuracy. The secondary structure assignment in the center of transglobular connections is correct in more than 85% of the cases. A similar error rate is found for loops. Since more global information about the fold is provided, it is complementary to standard secondary structure prediction approaches. Consequently, it may be useful in early stages of tertiary structure prediction when establishment of the structural class and possible folding topologies is of interest.

Does a backwardly read protein sequence have a unique native state?

Amino acid sequences of native proteins are generally not palindromic. Nevertheless, the protein molecule obtained as a result of reading the sequence backwards, i.e. a retro-protein, obviously has the same amino acid composition and the same hydrophobicity profile as the native sequence. The important questions which arise in the context of retro-proteins are: does a retro-protein fold to a well defined native-like structure as natural proteins do and, if the answer is positive, does a retro-protein fold to a structure similar to the native conformation of the original protein? In this work, the fold of retro-protein A, originated from the retro-sequence of the B domain of Staphylococcal protein A, was studied. As a result of lattice model simulations, it is conjectured that the retro-protein A also forms a three-helix bundle structure in solution. It is also predicted that the topology of the retro-protein A three-helix bundle is that of the native protein A, rather than that corresponding to the mirror image of native protein A. Secondary structure elements in the retro-protein do not exactly match their counterparts in the original protein structure; however, the amino acid side chain contract pattern of the hydrophobic core is partly conserved.

Prediction of quaternary structure of coiled coils. Application to mutants of the GCN4 leucine zipper.

Using a simplified protein model, the equilibrium between different oligomeric species of the wild-type GCN4 leucine zipper and seven of its mutants have been predicted. Over the entire experimental concentration range, agreement with experiment is found in five cases, while in two cases agreement is found over a portion of the concentration range. These studies demonstrate a methodology for predicting coiled coil quaternary structure and allow for the dissection of the interactions responsible for the global fold. In agreement with the conclusion of Harbury et al., the results of the simulations indicate that the pattern of hydrophobic and hydrophilic residues alone is insufficient to define a protein's three-dimensional structure. In addition, these simulations indicate that the degree of chain association is determined by the balance between specific side-chain packing preferences and the entropy reduction associated with side-chain burial in higher-order multimers.