Protein fold recognition by total alignment probability

We present a protein fold-recognition method that uses a comprehensive statistical interpretation of structural Hidden Markov Models (HMMs). The structure/fold recognition is done by summing the probabilities of all sequence-to-structure alignments. The optimal alignment can be defined as the most probable, but suboptimal alignments may have comparable probabilities. These suboptimal alignments can be interpreted as optimal alignments to the "other" structures from the ensemble or optimal alignments under minor fluctuations in the scoring function. Summing probabilities for all alignments gives a complete estimate of sequence-model compatibility. In the case of HMMs that produce a sequence, this reflects the fact that due to our indifference to exactly how the HMM produced the sequence, we should sum over all possibilities. We have built a set of structural HMMs for 188 protein structures and have compared two methods for identifying the structure compatible with a sequence: by the optimal alignment probability and by the total probability. Fold recognition by total probability was 40% more accurate than fold recognition by the optimal alignment probability. Proteins 2000;40:451-462.     

Protein fold recognition score functions: unusual construction strategies.

We describe two ways of optimizing score functions for protein sequence to structure threading. The first method adjusts parameters to improve sequence to structure alignment. The second adjusts parameters so as to improve a score function's ability to rank alignments calculated in the first score function. Unlike those functions known as knowledge-based force fields, the resulting parameter sets do not rely on Boltzmann statistics, have no claim to representing free energies and are purely constructions for recognizing protein folds. The methods give a small improvement, but suggest that functions can be profitably optimized for very specific aspects of protein fold recognition. Proteins 1999;36:454-461.     

A study on protein sequence alignment quality

One of the most central methods in bioinformatics is the alignment of two protein or DNA sequences. However, so far large-scale benchmarks examining the quality of these alignments are scarce. On the other hand, recently several large-scale studies of the capacity of different methods to identify related sequences has led to new insights about the performance of fold recognition methods. To increase our understanding about fold recognition methods, we present a large-scale benchmark of alignment quality. We compare alignments from several different alignment methods, including sequence alignments, hidden Markov models, PSI-BLAST, CLUSTALW, and threading methods. For most methods, the alignment quality increases significantly at about 20% sequence identity. The difference in alignment quality between different methods is quite small, and the main difference can be seen at the exact positioning of the sharp rise in alignment quality, that is, around 15-20% sequence identity. The alignments are improved by using structural information. In general, the best alignments are obtained by methods that use predicted secondary structure information and sequence profiles obtained from PSI-BLAST. One interesting observation is that for different pairs many different methods create the best alignments. This finding implies that if a method that could select the best alignment method for each pair existed, a significant improvement of the alignment quality could be gained.     

A fast, stochastic threading algorithm for proteins

MOTIVATION: Sequences for new proteins are being determined at a rapid rate, as a result of the Human Genome Project, and related genome research. The ability to predict the three-dimensional structure of proteins from sequence alone would be useful in discovering and understanding their function. Threading, or fold recognition, aims to predict the tertiary structure of a protein by aligning its amino acid sequence with a large number of structures, and finding the best fit. This approach depends on obtaining good performance from both the scoring function, which simulates the free energy for given trial alignments, and the threading algorithm, which searches for the lowest-score alignment. It appears that current scoring functions and threading algorithms need improvement. RESULTS: This paper presents a new threading algorithm. Numerical tests demonstrate that it is more powerful than two popular approximate algorithms, and much faster than exact methods.     

Consistency matrices: quantified structure alignments for sets of related proteins

Comparing two remotely similar structures is a difficult problem: more often than not, resulting structure alignments will show ambiguities and a unique answer usually does not even exist. In addition, alignments in general have a limited information content because every aligned residue is considered equally important. To solve these issues to a certain extent, one can take the perspective of a whole group of similar structures and then evaluate common structural features. Here, we describe a consistency approach that, although not actually performing a multiple structure alignment, does produce the information that one would conceivably want from such an experiment: the key structural features of the group, e.g., a fold, which in this case are projected onto either a pair of proteins or a single protein. Both representations are useful for a number of applications, ranging from the detection of (partially) wrong structure alignments to protein structure classification and fold recognition. To demonstrate some of these applications, the procedure was applied to 195 SCOP folds containing a total of 1802 domains sharing very low sequence similarity.     

Determining the roles of different chain fragments in recognition of immunoglobulin fold

We examine sequence-to-structure specificity of beta-structural fragments of immunoglobulin domains. The structure specificity of separate chain fragments is estimated by computing the Z-score values in recognition of the native structure in gapless threading tests. To improve the accuracy of our calculations we use energy averaging over diverse homologs of immunoglobulin domains. We show that the interactions between residues of beta-structure are more determinant in recognition of the native structure than the interactions within the whole chain molecule. This result distinguishes immunoglobulins from more typical proteins where the interactions between residues of the whole chain normally recognize the native fold more accurately than interactions between the residues of the secondary structure residues alone [Reva,B. and Topiol,S. (2000) BIOCOMPUTING: Proceedings of the Pacific Symposium. World Scientific Publishing Co., pp. 168-178]. We also find that the predominant contributions of the secondary structure are produced by the four central beta-strands that form the core of the molecule. The results of this study allow us through quantitative means to understand the architecture of immunoglobulin molecules. Comparing the fold recognition data for different chain fragments one can say that beta-strands form a rigid frame for immunoglobulin molecules, whereas loops, with no structural role, can develop a broad variety of binding specificities. It is well known that protein function is determined by specific portions of a protein chain. This study suggests that the whole protein structure can be predominantly determined by a few fragments of chain which form the structural framework of the molecule. This idea may help in better understanding the mechanisms of protein evolution: strengthening a protein structure in the key framework-forming regions allows mutations and flexibility in other chain regions.     

A method for the improvement of threading-based protein models

A new method for the homology-based modeling of protein three-dimensional structures is proposed and evaluated. The alignment of a query sequence to a structural template produced by threading algorithms usually produces low-resolution molecular models. The proposed method attempts to improve these models. In the first stage, a high-coordination lattice approximation of the query protein fold is built by suitable tracking of the incomplete alignment of the structural template and connection of the alignment gaps. These initial lattice folds are very similar to the structures resulting from standard molecular modeling protocols. Then, a Monte Carlo simulated annealing procedure is used to refine the initial structure. The process is controlled by the model's internal force field and a set of loosely defined restraints that keep the lattice chain in the vicinity of the template conformation. The internal force field consists of several knowledge-based statistical potentials that are enhanced by a proper analysis of multiple sequence alignments. The template restraints are implemented such that the model chain can slide along the template structure or even ignore a substantial fraction of the initial alignment. The resulting lattice models are, in most cases, closer (sometimes much closer) to the target structure than the initial threading-based models. All atom models could easily be built from the lattice chains. The method is illustrated on 12 examples of target/template pairs whose initial threading alignments are of varying quality. Possible applications of the proposed method for use in protein function annotation are briefly discussed.     

Using Evolutionary Trees in Protein Secondary Structure Prediction and Other Comparative Sequence Analyses

Previously proposed methods for protein secondary structure prediction from multiple sequence alignments do not efficiently extract the evolutionary information that these alignments contain. The predictions of these methods are less accurate than they could be, because of their failure to consider explicitly the phylogenetic tree that relates aligned protein sequences. As an alternative, we present a hidden Markov model approach to secondary structure prediction that more fully uses the evolutionary information contained in protein sequence alignments. A representative example is presented, and three experiments are performed that illustrate how the appropriate representation of evolutionary relatedness can improve inferences. We explain why similar improvement can be expected in other secondary structure prediction methods and indeed any comparative sequence analysis method.     

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.     

A multiple-template approach to protein threading

Most threading methods predict the structure of a protein using only a single template. Due to the increasing number of solved structures, a protein without solved structure is very likely to have more than one similar template structures. Therefore, a natural question to ask is if we can improve modeling accuracy using multiple templates. This article describes a new multiple-template threading method to answer this question. At the heart of this multiple-template threading method is a novel probabilistic-consistency algorithm that can accurately align a single protein sequence simultaneously to multiple templates. Experimental results indicate that our multiple-template method can improve pairwise sequence-template alignment accuracy and generate models with better quality than single-template models even if they are built from the best single templates (P-value <10(-6)) while many popular multiple sequence/structure alignment tools fail to do so. The underlying reason is that our probabilistic-consistency algorithm can generate accurate multiple sequence/template alignments. In another word, without an accurate multiple sequence/template alignment, the modeling accuracy cannot be improved by simply using multiple templates to increase alignment coverage. Blindly tested on the CASP9 targets with more than one good template structures, our method outperforms all other CASP9 servers except two (Zhang-Server and QUARK of the same group). Our probabilistic-consistency algorithm can possibly be extended to align multiple protein/RNA sequences and structures.     

3DCoffee: combining protein sequences and structures within multiple sequence alignments

Most bioinformatics analyses require the assembly of a multiple sequence alignment. It has long been suspected that structural information can help to improve the quality of these alignments, yet the effect of combining sequences and structures has not been evaluated systematically. We developed 3DCoffee, a novel method for combining protein sequences and structures in order to generate high-quality multiple sequence alignments. 3DCoffee is based on TCoffee version 2.00, and uses a mixture of pairwise sequence alignments and pairwise structure comparison methods to generate multiple sequence alignments. We benchmarked 3DCoffee using a subset of HOMSTRAD, the collection of reference structural alignments. We found that combining TCoffee with the threading program Fugue makes it possible to improve the accuracy of our HOMSTRAD dataset by four percentage points when using one structure only per dataset. Using two structures yields an improvement of ten percentage points. The measures carried out on HOM39, a HOMSTRAD subset composed of distantly related sequences, show a linear correlation between multiple sequence alignment accuracy and the ratio of number of provided structure to total number of sequences. Our results suggest that in the case of distantly related sequences, a single structure may not be enough for computing an accurate multiple sequence alignment.     

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.     

The PredictProtein server

PredictProtein (PP, http://cubic.bioc.columbia.edu/pp/) is an internet service for sequence analysis and the prediction of aspects of protein structure and function. Users submit protein sequence or alignments; the server returns a multiple sequence alignment, PROSITE sequence motifs, low-complexity regions (SEG), ProDom domain assignments, nuclear localisation signals, regions lacking regular structure and predictions of secondary structure, solvent accessibility, globular regions, transmembrane helices, coiled-coil regions, structural switch regions and disulfide-bonds. Upon request, fold recognition by prediction-based threading is available. For all services, users can submit their query either by electronic mail or interactively from World Wide Web.     

AL2CO: calculation of positional conservation in a protein sequence alignment

MOTIVATION: Amino acid sequence alignments are widely used in the analysis of protein structure, function and evolutionary relationships. Proteins within a superfamily usually share the same fold and possess related functions. These structural and functional constraints are reflected in the alignment conservation patterns. Positions of functional and/or structural importance tend to be more conserved. Conserved positions are usually clustered in distinct motifs surrounded by sequence segments of low conservation. Poorly conserved regions might also arise from the imperfections in multiple alignment algorithms and thus indicate possible alignment errors. Quantification of conservation by attributing a conservation index to each aligned position makes motif detection more convenient. Mapping these conservation indices onto a protein spatial structure helps to visualize spatial conservation features of the molecule and to predict functionally and/or structurally important sites. Analysis of conservation indices could be a useful tool in detection of potentially misaligned regions and will aid in improvement of multiple alignments. RESULTS: We developed a program to calculate a conservation index at each position in a multiple sequence alignment using several methods. Namely, amino acid frequencies at each position are estimated and the conservation index is calculated from these frequencies. We utilize both unweighted frequencies and frequencies weighted using two different strategies. Three conceptually different approaches (entropy-based, variance-based and matrix score-based) are implemented in the algorithm to define the conservation index. Calculating conservation indices for 35522 positions in 284 alignments from SMART database we demonstrate that different methods result in highly correlated (correlation coefficient more than 0.85) conservation indices. Conservation indices show statistically significant correlation between sequentially adjacent positions i and i + j, where j < 13, and averaging of the indices over the window of three positions is optimal for motif detection. Positions with gaps display substantially lower conservation properties. We compare conservation properties of the SMART alignments or FSSP structural alignments to those of the ClustalW alignments. The results suggest that conservation indices should be a valuable tool of alignment quality assessment and might be used as an objective function for refinement of multiple alignments. AVAILABILITY: The C code of the AL2CO program and its pre-compiled versions for several platforms as well as the details of the analysis are freely available at ftp://iole.swmed.edu/pub/al2co/.     

Large-scale comparison of protein sequence alignment algorithms with structure alignments

Sequence alignment programs such as BLAST and PSI-BLAST are used routinely in pairwise, profile-based, or intermediate-sequence-search (ISS) methods to detect remote homologies for the purposes of fold assignment and comparative modeling. Yet, the sequence alignment quality of these methods at low sequence identity is not known. We have used the CE structure alignment program (Shindyalov and Bourne, Prot Eng 1998;11:739) to derive sequence alignments for all superfamily and family-level related proteins in the SCOP domain database. CE aligns structures and their sequences based on distances within each protein, rather than on interprotein distances. We compared BLAST, PSI-BLAST, CLUSTALW, and ISS alignments with the CE structural alignments. We found that global alignments with CLUSTALW were very poor at low sequence identity (<25%), as judged by the CE alignments. We used PSI-BLAST to search the nonredundant sequence database (nr) with every sequence in SCOP using up to four iterations. The resulting matrix was used to search a database of SCOP sequences. PSI-BLAST is only slightly better than BLAST in alignment accuracy on a per-residue basis, but PSI-BLAST matrix alignments are much longer than BLAST's, and so align correctly a larger fraction of the total number of aligned residues in the structure alignments. Any two SCOP sequences in the same superfamily that shared a hit or hits in the nr PSI-BLAST searches were identified as linked by the shared intermediate sequence. We examined the quality of the longest SCOP-query/ SCOP-hit alignment via an intermediate sequence, and found that ISS produced longer alignments than PSI-BLAST searches alone, of nearly comparable per-residue quality. At 10-15% sequence identity, BLAST correctly aligns 28%, PSI-BLAST 40%, and ISS 46% of residues according to the structure alignments. We also compared CE structure alignments with FSSP structure alignments generated by the DALI program. In contrast to the sequence methods, CE and structure alignments from the FSSP database identically align 75% of residue pairs at the 10-15% level of sequence identity, indicating that there is substantial room for improvement in these sequence alignment methods. BLAST produced alignments for 8% of the 10,665 nonimmunoglobulin SCOP superfamily sequence pairs (nearly all <25% sequence identity), PSI-BLAST matched 17% and the double-PSI-BLAST ISS method aligned 38% with E-values <10.0. The results indicate that intermediate sequences may be useful not only in fold assignment but also in achieving more complete sequence alignments for comparative modeling.     

A simple iterative approach to parameter optimization.

Various bioinformatics problems require optimizing several different properties simultaneously. For example, in the protein threading problem, a scoring function combines the values for different parameters of possible sequence-to-structure alignments into a single score to allow for unambiguous optimization. In this context, an essential question is how each property should be weighted. As the native structures are known for some sequences, a partial ordering on optimal alignments to other structures, e.g., derived from structural comparisons, may be used to adjust the weights. To resolve the arising interdependence of weights and computed solutions, we propose a heuristic approach: iterating the computation of solutions (here, threading alignments) given the weights and the estimation of optimal weights of the scoring function given these solutions via systematic calibration methods. For our application (i.e., threading), this iterative approach results in structurally meaningful weights that significantly improve performance on both the training and the test data sets. In addition, the optimized parameters show significant improvements on the recognition rate for a grossly enlarged comprehensive benchmark, a modified recognition protocol as well as modified alignment types (local instead of global and profiles instead of single sequences). These results show the general validity of the optimized weights for the given threading program and the associated scoring contributions.     

Analogy-based protein structure prediction: II. Testing of substitution matrices and pseudopotentials used to align protein sequences with spatial structures

The results of testing the recognition ability of various amino acid substitution matrices and manifold (both extracted from the literature and of our own design) pseudopotentials intended for the recognition of protein structures and sequence-to-structure alignments are described. The numerical estimates of the recognition ability of various substitution matrices and pseudopotentials were obtained for different levels of protein structure similarity. It is demonstrated that substitution matrices work much better than pseudopotentials at a high degree of sequence similarity of spatially similar proteins; however, some pseudopotentials outdo substitution matrices at a low level of sequence similarity between analogous proteins.     

Fold prediction by a hierarchy of sequence, threading, and modeling methods.

Several fold recognition algorithms are compared to each other in terms of prediction accuracy and significance. It is shown that on standard benchmarks, hybrid methods, which combine scoring based on sequence-sequence and sequence-structure matching, surpass both sequence and threading methods in the number of accurate predictions. However, the sequence similarity contributes most to the prediction accuracy. This strongly argues that most examples of apparently nonhomologous proteins with similar folds are actually related by evolution. While disappointing from the perspective of the fundamental understanding of protein folding, this adds a new significance to fold recognition methods as a possible first step in function prediction. Despite hybrid methods being more accurate at fold prediction than either the sequence or threading methods, each of the methods is correct in some cases where others have failed. This partly reflects a different perspective on sequence/structure relationship embedded in various methods. To combine predictions from different methods, estimates of significance of predictions are made for all methods. With the help of such estimates, it is possible to develop a "jury" method, which has accuracy higher than any of the single methods. Finally, building full three-dimensional models for all top predictions helps to eliminate possible false positives where alignments, which are optimal in the one-dimensional sequences, lead to unsolvable sterical conflicts for the full three-dimensional models.