Protein structure prediction by threading methods: Evaluation of current techniques

This paper evaluates the results of a protein structure prediction contest. The predictions were made using threading procedures, which employ techniques for aligning sequences with 3D structures to select the correct fold of a given sequence from a set of alternatives. Nine different teams submitted 86 predictions, on a total of 21 target proteins with little or no sequence homology to proteins of known structure. The 3D structures of these proteins were newly determined by experimental methods, but not yet published or otherwise available to the predictors. The predictions, made from the amino acid sequence alone, thus represent a genuine test of the current performance of threading methods. Only a subset of all the predictions is evaluated here. It corresponds to the 44 predictions submitted for the 11 target proteins seen to adopt known folds. The predictions for the remaining 10 proteins were not analyzed, although weak similarities with known folds may also exist in these proteins. We find that threading methods are capable of identifying the correct fold in many cases, but not reliably enough as yet. Every team predicts correctly a different set of targets, with virtually all targets predicted correctly by at least one team. Also, common folds such as TIM barrels are recognized more readily than folds with only a few known examples. However, quite surprisingly, the quality of the sequence-structure alignments, corresponding to correctly recognized folds, is generally very poor, as judged by comparison with the corresponding 3D structure alignments. Thus, threading can presently not be relied upon to derive a detailed 3D model from the amino acid sequence. This raises a very intriguing question: how is fold recognition achieved? Our analysis suggests that it may be achieved because threading procedures maximize hydrophobic interactions in the protein core, and are reasonably good at recognizing local secondary structure. © 1995 Wiley-Liss, Inc.     

3D-SHOTGUN: a novel,cooperative, fold-recognition meta-predictor

To gain a better understanding of the biological role of proteins encoded in genome sequences, knowledge of their three-dimensional (3D) structure and function is required. The computational assignment of folds is becoming an increasingly important complement to experimental structure determination. In particular, fold-recognition methods aim to predict approximate 3D models for proteins bearing no sequence similarity to any protein of known structure. However, fully automated structure-prediction methods can currently produce reliable models for only a fraction of these sequences. Using a number of semiautomated procedures, human expert predictors are often able to produce more and better predictions than automated methods. We describe a novel, fully automatic, fold-recognition meta-predictor, named 3D-SHOTGUN, which incorporates some of the strategies human predictors have successfully applied. This new method is reminiscent of the so-called cooperative algorithms of Computer Vision. The input to 3D-SHOTGUN are the top models predicted by a number of independent fold-recognition servers. The meta-predictor consists of three steps: (i) assembly of hybrid models, (ii) confidence assignment, and (iii) selection. We have applied 3D-SHOTGUN to an unbiased test set of 77 newly released protein structures sharing no sequence similarity to proteins previously released. Forty-six correct rank-1 predictions were obtained, 30 of which had scores higher than that of the first incorrect prediction-a significant improvement over the performance of all individual servers. Furthermore, the predicted hybrid models were, on average, more similar to their corresponding native structures than those produced by the individual servers. This opens the possibility of generating more accurate, full-atom homology models for proteins with no sequence similarity to proteins of known structure. These improvements represent a step forward toward the wider applicability of fully automated structure-prediction methods at genome scales.     

Threading a database of protein cores

We present an analysis of 10 blind predictions prepared for a recent conference, “Critical Assessment of Techniques for Protein Structure Prediction.”¹ The sequences of these proteins are not detectably similar to those of any protein in the structure database then available, but we attempted, by a threading method, to recognize similarity to known domain folds. Four of the 10 proteins, as we subsequently learned, do indeed show significant similarity to then-known structures. For 2 of these proteins the predictions were accurate, in the sense that a similar structure was at or near the top of the list of threading scores, and the threading alignment agreed well with the corresponding structural alignment. For the best predicted model mean alignment error relative to the optimal structural alignment was 2.7 residues, arising entirely from small “register shifts” of strands or helices. In the analysis we attempt to identify factors responsible for these successes and failures. Since our threading method does not use gap penalties, we may readily distinguish between errors arising from our prior definition of the “cores” of known structures and errors arising from inherent limitations in the threading potential. It would appear from the results that successful substructure recognition depends most critically on accurate definition of the “fold” of a database protein. This definition must correctly delineate substructures that are, and are not, likely to be conserved during protein evolution. © 1995 Wiley-Liss, Inc.     

Progress in fold recognition

The prediction experiment reveals that fold recognition has become a powerful tool in structural biology. We applied our fold recognition technique to 13 target sequences. In two cases, replication terminating protein and prosequence of subtilisin, the predicted structures are very similar to the experimentally determined folds. For the first time, in a public blind test, the unknown structures of proteins have been predicted ahead of experiment to an accuracy approaching molecular detail. In two other cases the approximate folds have been predicted correctly. According to the assessors there were 12 recognizable folds among the target proteins. In our postprediction analysis we find that in 7 cases our fold recognition technique is successful. In several of the remaining cases the predicted folds have interesting features in common with the experimental results. We present our procedure, discuss the results, and comment on several fundamental and technical problems encountered in fold recognition. © 1995 Wiley-Liss, Inc.     

A method for α-helical integral membrane protein fold prediction

Integral membrane proteins (of the α-helical class) are of central importance in a wide variety of vital cellular functions. Despite considerable effort on methods to predict the location of the helices, little attention has been directed toward developing an automatic method to pack the helices together. In principle, the prediction of membrane proteins should be easier than the prediction of globular proteins: there is only one type of secondary structure and all helices pack with a common alignment across the membrane. This allows all possible structures to be represented on a simple lattice and exhaustively enumerated. Prediction success lies not in generating many possible folds but in recognizing which corresponds to the native. Our evaluation of each fold is based on how well the exposed surface predicted from a multiple sequence alignment fits its allocated position. Just as exposure to solvent in globular proteins can be predicted from sequence variation, so exposure to lipid can be recognized by variable-hydrophobic (variphobic) positions. Application to both bacteriorhodopsin and the eukaryotic rhodopsin/opsin families revealed that the angular size of the lipid-exposed faces must be predicted accurately to allow selection of the correct fold. With the inherent uncertainties in helix prediction and parameter choice, this accuracy could not be guaranteed but the correct fold was typically found in the top six candidates. Our method provides the first completely automatic method that can proceed from a scan of the protein sequence databanks to a predicted three-dimensional structure with no intervention required from the investigator. Within the limited domain of the seven helix bundle proteins, a good chance can be given of selecting the correct structure. However, the limited number of sequences available with a corresponding known structure makes further characterization of the method difficult. © 1994 John Wiley & Sons, Inc.     

Evaluation of current techniques for Ab initio protein structure prediction

The results of a protein structure prediction contest are reviewed. Twelve different groups entered predictions on 14 proteins of known sequence whose structures had been determined but not yet disseminated to the scientific community. Thus, these represent true tests of the current state of structure prediction methodologies. From this work, it is clear that accurate tertiary structure prediction is not yet possible. However, protein fold and motif prediction are possible when the motif is recognizably similar to another known structure. Internal symmetry and the information inherent in an aligned family of homologous sequences facilitate predictive efforts. Novel folds remain a major challenge for prediction efforts. © 1995 Wiley-Liss, Inc.     

Targeting novel folds for structural genomics

The ultimate goal of structural genomics is to obtain the structure of each protein coded by each gene within a genome to determine gene function. Because of cost and time limitations, it remains impractical to solve the structure for every gene product experimentally. Up to a point, reasonably accurate three‐dimensional structures can be deduced for proteins with homologous sequences by using comparative modeling. Beyond this, fold recognition or threading methods can be used for proteins showing little homology to any known fold, although this is relatively time‐consuming and limited by the library of template folds currently available. Therefore, it is appropriate to develop methods that can increase our knowledge base, expanding our fold libraries by earmarking potentially “novel” folds for experimental structure determination. How can we sift through proteomic data rapidly and yet reliably identify novel folds as targets for structural genomics? We have analyzed a number of simple methods that discriminate between “novel” and “known” folds. We propose that simple alignments of secondary structure elements using predicted secondary structure could potentially be a more selective method than both a simple fold recognition method (GenTHREADER) and standard sequence alignment at finding novel folds when sequences show no detectable homology to proteins with known structures. Proteins 2002;48:44–52. © 2002 Wiley‐Liss, Inc.     

Protein fold recognition using sequence-derived predictions.

In protein fold recognition, one assigns a probe amino acid sequence of unknown structure to one of a library of target 3D structures. Correct assignment depends on effective scoring of the probe sequence for its compatibility with each of the target structures. Here we show that, in addition to the amino acid sequence of the probe, sequence-derived properties of the probe sequence (such as the predicted secondary structure) are useful in fold assignment. The additional measure of compatibility between probe and target is the level of agreement between the predicted secondary structure of the probe and the known secondary structure of the target fold. That is, we recommend a sequence-structure compatibility function that combines previously developed compatibility functions (such as the 3D-1D scores of Bowie et al. [1991] or sequence-sequence replacement tables) with the predicted secondary structure of the probe sequence. The effect on fold assignment of adding predicted secondary structure is evaluated here by using a benchmark set of proteins (Fischer et al., 1996a). The 3D structures of the probe sequences of the benchmark are actually known, but are ignored by our method. The results show that the inclusion of the predicted secondary structure improves fold assignment by about 25%. The results also show that, if the true secondary structure of the probe were known, correct fold assignment would increase by an additional 8-32%. We conclude that incorporating sequence-derived predictions significantly improves assignment of sequences to known 3D folds. Finally, we apply the new method to assign folds to sequences in the SWISSPROT database; six fold assignments are given that are not detectable by standard sequence-sequence comparison methods; for two of these, the fold is known from X-ray crystallography and the fold assignment is correct.     

Ab initio prediction of the three-dimensional structure of a de novo designed protein: a double-blind case study

Ab initio structure prediction and de novo protein design are two problems at the forefront of research in the fields of structural biology and chemistry. The goal of ab initio structure prediction of proteins is to correctly characterize the 3D structure of a protein using only the amino acid sequence as input. De novo protein design involves the production of novel protein sequences that adopt a desired fold. In this work, the results of a double-blind study are presented in which a new ab initio method was successfully used to predict the 3D structure of a protein designed through an experimental approach using binary patterned combinatorial libraries of de novo sequences. The predicted structure, which was produced before the experimental structure was known and without consideration of the design goals, and the final NMR analysis both characterize this protein as a 4-helix bundle. The similarity of these structures is evidenced by both small RMSD values between the coordinates of the two structures and a detailed analysis of the helical packing.     

Protein structural similarities predicted by a sequence-structure compatibility method.

A method for protein structure prediction has been developed, which evaluates the compatibility of an amino acid sequence with known 3-dimensional structures and identifies the most likely structure. The method was applied to a large number of sequences in a database, and the structures of the following proteins were predicted: (1) shikimate kinase (SKase), (2) the hydrophilic subunit of mannose permease (IIABMan), (3) rat tyrosine aminotransferase (Tyr AT), and (4) threonine dehydratase (TDH). The functional and evolutionary implications of the predictions are discussed. (1) The structural similarity between SKase and adenylate kinase was predicted. Alignment of their sequences reveals that the ATP-binding type A sequence motif and 2 ATP-binding arginine residues are conserved. The prediction suggests a similarity in their functional mechanisms as well as an evolutionary relationship. (2) The structural similarity between IIABMan and galactose/glucose-binding protein (GGBP) was predicted. The IIA and IIB domains are aligned with the N- and C-terminal domains of GGBP, respectively. The 2 phosphorylated residues, His 10 and His 175, of IIABMan are threaded onto loops located in the substrate-binding cleft of GGBP. The prediction accounts for the phosphoryl transfer from His 10 to His 175, and to the sugar substrate. (3) The structural similarity between rat Tyr AT and Escherichia coli aspartate AT was predicted, as well as (4) the structural similarity between TDH and the tryptophan synthase beta subunit. Predictions (3) and (4) support the previous predictions based on observations of the functional similarities between the proteins.     

Secondary structure prediction and unrefined tertiary structure prediction for cyclin A,B, and D

We present heuristic-based predictions of the secondary and tertiary structures of the cyclins A, B, and D, representatives of the cyclin superfamily. The list of suggested constraints for tertiary structure assembly was left unrefined in order to submit this report before an announced crystal structure for cyclin A becomes available. To predict these constraints, a master sequence alignment over 270 positions of cyclin types A, B, and D was adjusted based on individual secondary structure predictions for each type. We used new heuristics for predicting aromatic residues at protein-protein interfaces and to identify sequentially distinct regions in the protein chain that cluster in the folded structure. The boundaries of two conjectured domains in the cyclin fold were predicted based on experimental data in the literature. The domain that is important for interaction of the cyclins with cyclin-dependent kinases (CDKs) is predicted to contain six helices; the second domain in the consensus model contains both helices and a β-sheet that is formed by sequentially distant regions in the protein chain. A plausible phosphorylation site is identified. This work represents a blinded test of the method for prediction of secondary and, to a lesser extent, tertiary structure from a set of homologous protein sequences. Evaluation of our predictions will become possible with the publication of the announced crystal structure.     

Modeling repetitive,non‐globular proteins

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

Fold recognition and accurate sequence-structure alignment of sequences directing beta-sheet proteins

The ability to predict structure from sequence is particularly important for toxins, virulence factors, allergens, cytokines, and other proteins of public health importance. Many such functions are represented in the parallel beta-helix and beta-trefoil families. A method using pairwise beta-strand interaction probabilities coupled with evolutionary information represented by sequence profiles is developed to tackle these problems for the beta-helix and beta-trefoil folds. The algorithm BetaWrapPro employs a "wrapping" component that may capture folding processes with an initiation stage followed by processive interaction of the sequence with the already-formed motifs. BetaWrapPro outperforms all previous motif recognition programs for these folds, recognizing the beta-helix with 100% sensitivity and 99.7% specificity and the beta-trefoil with 100% sensitivity and 92.5% specificity, in crossvalidation on a database of all nonredundant known positive and negative examples of these fold classes in the PDB. It additionally aligns 88% of residues for the beta-helices and 86% for the beta-trefoils accurately (within four residues of the exact position) to the structural template, which is then used with the side-chain packing program SCWRL to produce 3D structure predictions. One striking result has been the prediction of an unexpected parallel beta-helix structure for a pollen allergen, and its recent confirmation through solution of its structure. A Web server running BetaWrapPro is available and outputs putative PDB-style coordinates for sequences predicted to form the target folds.     

Large scale ab initio modeling of structurally uncharacterized antimicrobial peptides reveals known and novel folds

Antimicrobial resistance within a wide range of infectious agents is a severe and growing public health threat. Antimicrobial peptides (AMPs) are among the leading alternatives to current antibiotics, exhibiting broad spectrum activity. Their activity is determined by numerous properties such as cationic charge, amphipathicity, size, and amino acid composition. Currently, only around 10% of known AMP sequences have experimentally solved structures. To improve our understanding of the AMP structural universe we have carried out large scale ab initio 3D modeling of structurally uncharacterized AMPs that revealed similarities between predicted folds of the modeled sequences and structures of characterized AMPs. Two of the peptides whose models matched known folds are Lebocin Peptide 1A (LP1A) and Odorranain M, predicted to form β-hairpins but, interestingly, to lack the intramolecular disulfide bonds, cation-π or aromatic interactions that generally stabilize such AMP structures. Other examples include Ponericin Q42, Latarcin 4a, Kassinatuerin 1, Ceratotoxin D, and CPF-B1 peptide, which have α-helical folds, as well as mixed αβ folds of human Histatin 2 peptide and Garvicin A which are, to the best of our knowledge, the first linear αββ fold AMPs lacking intramolecular disulfide bonds. In addition to fold matches to experimentally derived structures, unique folds were also obtained, namely for Microcin M and Ipomicin. These results help in understanding the range of protein scaffolds that naturally bear antimicrobial activity and may facilitate protein design efforts towards better AMPs.     

Efficient identification of near‐native conformations in ab initio protein structure prediction using structural profiles

One of the major bottlenecks in many ab initio protein structure prediction methods is currently the selection of a small number of candidate structures for high‐resolution refinement from large sets of low‐resolution decoys. This step often includes a scoring by low‐resolution energy functions and a clustering of conformations by their pairwise root mean square deviations (RMSDs). As an efficient selection is crucial to reduce the overall computational cost of the predictions, any improvement in this direction can increase the overall performance of the predictions and the range of protein structures that can be predicted. We show here that the use of structural profiles, which can be predicted with good accuracy from the amino acid sequences of proteins, provides an efficient means to identify good candidate structures. Proteins 2010. © 2009 Wiley‐Liss, Inc.     

Sequence and structure alignments in post-AlphaFold era

Sequence alignment is fundamental for analyzing protein structure and function. For all but closely-related proteins, alignments based on structures are more accurate than alignments based purely on amino-acid sequences. However, the disparity between the large amount of sequence data and the relative paucity of experimentally-determined structures has precluded the general applicability of structure alignment. Based on the success of AlphaFold (and its likes) in producing high-quality structure predictions, we suggest that when aligning homologous proteins, lacking experimental structures, better results can be obtained by a structural alignment of predicted structures than by an alignment based only on amino-acid sequences. We present a quantitative evaluation, based on pairwise alignments of sequences and structures (both predicted and experimental) to support this hypothesis.     

A test of proposed rules for helix capping: implications for protein design

alpha-helices within proteins are often terminated (capped) by distinctive configurations of the polypeptide chain. Two common arrangements are the Schellman motif and the alternative alpha(L) motif. Rose and coworkers developed stereochemical rules to identify the locations of such motifs in proteins of unknown structure based only on their amino acid sequences. To check the effectiveness of these rules, they made specific predictions regarding the structural and thermodynamic consequences of certain mutations in T4 lysozyme. We have constructed these mutants and show here that they have neither the structure nor the stability that was predicted. The results show the complexity of the protein-folding problem. Comparison of known protein structures may show that a characteristic sequence of amino acids (a sequence motif) corresponds to a conserved structural motif. In any particular protein, however, changes in other parts of the sequence may result in a different conformation. The structure is determined by sequence as a whole, not by parts considered in isolation.     

Improving fold recognition without folds

The most reliable way to align two proteins of unknown structure is through sequence-profile and profile-profile alignment methods. If the structure for one of the two is known, fold recognition methods outperform purely sequence-based alignments. Here, we introduced a novel method that aligns generalised sequence and predicted structure profiles. Using predicted 1D structure (secondary structure and solvent accessibility) significantly improved over sequence-only methods, both in terms of correctly recognising pairs of proteins with different sequences and similar structures and in terms of correctly aligning the pairs. The scores obtained by our generalised scoring matrix followed an extreme value distribution; this yielded accurate estimates of the statistical significance of our alignments. We found that mistakes in 1D structure predictions correlated between proteins from different sequence-structure families. The impact of this surprising result was that our method succeeded in significantly out-performing sequence-only methods even without explicitly using structural information from any of the two. Since AGAPE also outperformed established methods that rely on 3D information, we made it available through. If we solved the problem of CPU-time required to apply AGAPE on millions of proteins, our results could also impact everyday database searches.     

Prediction, conservation analysis, and structural characterization of mammalian mucin-type O-glycosylation sites

O-GalNAc-glycosylation is one of the main types of glycosylation in mammalian cells. No consensus recognition sequence for the O-glycosyltransferases is known, making prediction methods necessary to bridge the gap between the large number of known protein sequences and the small number of proteins experimentally investigated with regard to glycosylation status. From O-GLYCBASE a total of 86 mammalian proteins experimentally investigated for in vivo O-GalNAc sites were extracted. Mammalian protein homolog comparisons showed that a glycosylated serine or threonine is less likely to be precisely conserved than a nonglycosylated one. The Protein Data Bank was analyzed for structural information, and 12 glycosylated structures were obtained. All positive sites were found in coil or turn regions. A method for predicting the location for mucin-type glycosylation sites was trained using a neural network approach. The best overall network used as input amino acid composition, averaged surface accessibility predictions together with substitution matrix profile encoding of the sequence. To improve prediction on isolated (single) sites, networks were trained on isolated sites only. The final method combines predictions from the best overall network and the best isolated site network; this prediction method correctly predicted 76% of the glycosylated residues and 93% of the nonglycosylated residues. NetOGlyc 3.1 can predict sites for completely new proteins without losing its performance. The fact that the sites could be predicted from averaged properties together with the fact that glycosylation sites are not precisely conserved indicates that mucin-type glycosylation in most cases is a bulk property and not a very site-specific one. NetOGlyc 3.1 is made available at www.cbs.dtu.dk/services/netoglyc.     

Cell surface proteins in archaeal and bacterial genomes comprising "LVIVD", "RIVW" and "LGxL" tandem sequence repeats are predicted to fold as beta-propeller

Proteins that share even low sequence homologies are known to adopt similar folds. The beta-propeller structural motif is one such example. Identifying sequences that adopt a beta-propeller fold is useful to annotate protein structure and function. Often, tandem sequence repeats provide the necessary signal for identifying beta-propellers in proteins. In our recent analysis to identify cell surface proteins in archaeal and bacterial genomes, we identified some proteins that contain novel tandem repeats "LVIVD", "RIVW" and "LGxL". In this work, based on protein fold predictions and three-dimensional comparative modeling methods, we predicted that these repeat types fold as beta-propeller. Further, the evolutionary trace analysis of all proteins constituting amino acid sequence repeats in beta-propellers suggest that the novel repeats have diverged from a common ancestor.