Cascaded multiple classifiers for secondary structure prediction

We describe a new classifier for protein secondary structure prediction that is formed by cascading together different types of classifiers using neural networks and linear discrimination. The new classifier achieves an accuracy of 76.7% (assessed by a rigorous full Jack-knife procedure) on a new nonredundant dataset of 496 nonhomologous sequences (obtained from G.J. Barton and J.A. Cuff). This database was especially designed to train and test protein secondary structure prediction methods, and it uses a more stringent definition of homologous sequence than in previous studies. We show that it is possible to design classifiers that can highly discriminate the three classes (H, E, C) with an accuracy of up to 78% for beta-strands, using only a local window and resampling techniques. This indicates that the importance of long-range interactions for the prediction of beta-strands has been probably previously overestimated.     

Improving the prediction of secondary structure of 'TIM-barrel' enzymes.

The information contained in aligned sets of homologous protein sequences should improve the score of secondary structure prediction. Seven different enzymes having the (beta/alpha)8 or TIM-barrel fold were used to optimize the prediction with regard to this class of enzymes. The alpha-helix, beta-strand and loop propensities of the Garnier-Osguthorpe-Robson method were averaged at aligned residue positions, leading to a significant improvement over the average score obtained from single sequences. The increased accuracy correlates with the average sequence variability of the aligned set. Further improvements were obtained by using the following averaged properties as weights for the averaged state propensities: amphipathic moment and alpha-helix; hydropathy and beta-strand; chain flexibility and loop. The clustering of conserved residues at the C-terminal ends of the beta-strands was used as an additional positive weight for beta-strand propensity and increased the prediction of otherwise unpredicted beta-strands decisively. The automatic weighted prediction method identifies greater than 95% of the secondary structure elements of the set of seven TIM-barrel enzymes.     

An algorithm for predicting protein-protein interaction sites: Abnormally exposed amino acid residues and secondary structure elements

Multiprotein systems mediate most regulatory processes in living organisms. Although the structures of the individual proteins are often defined, less is known of the structures of multiprotein systems. Computational methods for predicting interfaces, using evolutionary conservation and/or physicochemical data, have been developed. Here we consider the use of solvent accessibility, residue propensity, and hydrophobicity, in conjunction with secondary structure data, as prediction parameters. We analyze the influence of residue type and secondary structure on solvent accessibility and define a measure of "relative exposedness." Clustering abnormally high scoring residues provides a basis for predicting interaction sites. The analysis is extended to investigate abnormally exposed secondary structure elements, particularly beta-sheet strands. We show that surface-exposed beta-strands lacking protective features are more likely to be found at protein-protein interfaces, allowing us to create an algorithm with approximately 68% and approximately 75% accuracy in differentiating between interacting and edge strands in isolated beta-strands and beta-sheet strands, respectively. These methods of identifying abnormally exposed surface regions are combined in an algorithm, which, on a data set of 77 unbound and disjoint (single chain extracted from complex) structures, predicts 79% of the protein-protein interfaces correctly. If enzyme-inhibitor complexes, where the inhibitor mimics a nonprotein substrate, are excluded, the accuracy increases to 85%.     

Bayesian segmentation of protein secondary structure.

We present a novel method for predicting the secondary structure of a protein from its amino acid sequence. Most existing methods predict each position in turn based on a local window of residues, sliding this window along the length of the sequence. In contrast, we develop a probabilistic model of protein sequence/structure relationships in terms of structural segments, and formulate secondary structure prediction as a general Bayesian inference problem. A distinctive feature of our approach is the ability to develop explicit probabilistic models for alpha-helices, beta-strands, and other classes of secondary structure, incorporating experimentally and empirically observed aspects of protein structure such as helical capping signals, side chain correlations, and segment length distributions. Our model is Markovian in the segments, permitting efficient exact calculation of the posterior probability distribution over all possible segmentations of the sequence using dynamic programming. The optimal segmentation is computed and compared to a predictor based on marginal posterior modes, and the latter is shown to provide significant improvement in predictive accuracy. The marginalization procedure provides exact secondary structure probabilities at each sequence position, which are shown to be reliable estimates of prediction uncertainty. We apply this model to a database of 452 nonhomologous structures, achieving accuracies as high as the best currently available methods. We conclude by discussing an extension of this framework to model nonlocal interactions in protein structures, providing a possible direction for future improvements in secondary structure prediction accuracy.     

ChSeq: A database of chameleon sequences

Chameleon sequences (ChSeqs) refer to sequence strings of identical amino acids that can adopt different conformations in protein structures. Researchers have detected and studied ChSeqs to understand the interplay between local and global interactions in protein structure formation. The different secondary structures adopted by one ChSeq challenge sequence‐based secondary structure predictors. With increasing numbers of available Protein Data Bank structures, we here identify a large set of ChSeqs ranging from 6 to 10 residues in length. The homologous ChSeqs discovered highlight the structural plasticity involved in biological function. When compared with previous studies, the set of unrelated ChSeqs found represents an about 20‐fold increase in the number of detected sequences, as well as an increase in the longest ChSeq length from 8 to 10 residues. We applied secondary structure predictors on our ChSeqs and found that methods based on a sequence profile outperformed methods based on a single sequence. For the unrelated ChSeqs, the evolutionary information provided by the sequence profile typically allows successful prediction of the prevailing secondary structure adopted in each protein family. Our dataset will facilitate future studies of ChSeqs, as well as interpretations of the interplay between local and nonlocal interactions. A user‐friendly web interface for this ChSeq database is available at prodata.swmed.edu/chseq .     

Use of tetrapeptide signals for protein secondary-structure prediction

This paper develops a novel sequence-based method, tetra-peptide-based increment of diversity with quadratic discriminant analysis (TPIDQD for short), for protein secondary-structure prediction. The proposed TPIDQD method is based on tetra-peptide signals and is used to predict the structure of the central residue of a sequence fragment. The three-state overall per-residue accuracy (Q ₃) is about 80% in the threefold cross-validated test for 21-residue fragments in the CB513 dataset. The accuracy can be further improved by  taking long-range sequence information (fragments of more than 21 residues) into account in prediction. The results show the tetra-peptide signals can indeed reflect some relationship between an amino acid’s sequence and its secondary structure, indicating the importance of  tetra-peptide signals as the protein folding code in the protein structure prediction.     

What Is the Minimum Number of Residues to Determine the Secondary Structural State?

The failure of protein secondary structural prediction is commonly attributed to the neglect of long-range interactions. The question is, what is the minimum length of subsequence required to determine the central secondary structural state, stabilized only by local interactions? In the present work, the 20 amino acids were classified into eight groups to analyze systematically the relationship between the length and secondary structural state of subsequences in the PDB database. It was found that the fraction of subsequences with a unique central secondary structural state increases with increasing length, and the minimum length of subsequence required to determine the central secondary structural state is about 14–17 residues. The low accuracy of secondary structure prediction does not result from the neglect of long-range interactions, but may result from the limitation of the available protein database size or prediction algorithm.     

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

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

Use of secondary structural information and C<Emphasis Type="Italic">α</Emphasis>-C<Emphasis Type="Italic">α</Emphasis> distance restraints to model protein structures with MODELLER

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

On the Relation Between the Predicted Secondary Structure and the Protein Size

Accurately predicted protein secondary structure provides useful information for target selection, to analyze protein function and to predict higher dimensional structure. Existing research shows that more data + refined search = better prediction. We analyze relation between the prediction accuracy and another crucial factor, the protein size. Empirical tests performed with two secondary structure predictors on a large set of high-resolution, non-redundant proteins show that the average accuracies for small proteins (<100 residues) equal 73% and 54% for alpha-helices and beta-strands, respectively. The alpha-helix/beta-strand accuracies for very large proteins (>300 residues) equal 77%/68%, respectively. Similarly, the tests with three secondary structure content predictors show that the prediction errors for the small/very large proteins equal 0.13/0.09 and 0.09/0.06 for alpha-helix and beta-strand content, respectively. Our tests confirm that the secondary structure/content predictions for the very large proteins are characterized statistically significantly better quality than prediction for the small proteins. This is in contrast with the tertiary structure predictions in which higher accuracy is obtained for smaller proteins.     

Predicting residue-wise contact orders in proteins by support vector regression

Background  

The residue-wise contact order (RWCO) describes the sequence separations between the residues of interest and its contacting residues in a protein sequence. It is a new kind of one-dimensional protein structure that represents the extent of long-range contacts and is considered as a generalization of contact order. Together with secondary structure, accessible surface area, the B factor, and contact number, RWCO provides comprehensive and indispensable important information to reconstructing the protein three-dimensional structure from a set of one-dimensional structural properties. Accurately predicting RWCO values could have many important applications in protein three-dimensional structure prediction and protein folding rate prediction, and give deep insights into protein sequence-structure relationships.     

Identification and application of the concepts important for accurate and reliable protein secondary structure prediction.

A protein secondary structure prediction method from multiply aligned homologous sequences is presented with an overall per residue three-state accuracy of 70.1%. There are two aims: to obtain high accuracy by identification of a set of concepts important for prediction followed by use of linear statistics; and to provide insight into the folding process. The important concepts in secondary structure prediction are identified as: residue conformational propensities, sequence edge effects, moments of hydrophobicity, position of insertions and deletions in aligned homologous sequence, moments of conservation, auto-correlation, residue ratios, secondary structure feedback effects, and filtering. Explicit use of edge effects, moments of conservation, and auto-correlation are new to this paper. The relative importance of the concepts used in prediction was analyzed by stepwise addition of information and examination of weights in the discrimination function. The simple and explicit structure of the prediction allows the method to be reimplemented easily. The accuracy of a prediction is predictable a priori. This permits evaluation of the utility of the prediction: 10% of the chains predicted were identified correctly as having a mean accuracy of > 80%. Existing high-accuracy prediction methods are "black-box" predictors based on complex nonlinear statistics (e.g., neural networks in PHD: Rost & Sander, 1993a). For medium- to short-length chains (> or = 90 residues and < 170 residues), the prediction method is significantly more accurate (P < 0.01) than the PHD algorithm (probably the most commonly used algorithm). In combination with the PHD, an algorithm is formed that is significantly more accurate than either method, with an estimated overall three-state accuracy of 72.4%, the highest accuracy reported for any prediction method.     

Residue-level prediction of DNA-binding sites and its application on DNA-binding protein predictions

Protein-DNA interactions are crucial to many cellular activities such as expression-control and DNA-repair. These interactions between amino acids and nucleotides are highly specific and any aberrance at the binding site can render the interaction completely incompetent. In this study, we have three aims focusing on DNA-binding residues on the protein surface: to develop an automated approach for fast and reliable recognition of DNA-binding sites; to improve the prediction by distance-dependent refinement; use these predictions to identify DNA-binding proteins. We use a support vector machines (SVM)-based approach to harness the features of the DNA-binding residues to distinguish them from non-binding residues. Features used for distinction include the residue's identity, charge, solvent accessibility, average potential, the secondary structure it is embedded in, neighboring residues, and location in a cationic patch. These features collected from 50 proteins are used to train SVM. Testing is then performed on another set of 37 proteins, much larger than any testing set used in previous studies. The testing set has no more than 20% sequence identity not only among its pairs, but also with the proteins in the training set, thus removing any undesired redundancy due to homology. This set also has proteins with an unseen DNA-binding structural class not present in the training set. With the above features, an accuracy of 66% with balanced sensitivity and specificity is achieved without relying on homology or evolutionary information. We then develop a post-processing scheme to improve the prediction using the relative location of the predicted residues. Balanced success is then achieved with average sensitivity, specificity and accuracy pegged at 71.3%, 69.3% and 70.5%, respectively. Average net prediction is also around 70%. Finally, we show that the number of predicted DNA-binding residues can be used to differentiate DNA-binding proteins from non-DNA-binding proteins with an accuracy of 78%. Results presented here demonstrate that machine-learning can be applied to automated identification of DNA-binding residues and that the success rate can be ameliorated as more features are added. Such functional site prediction protocols can be useful in guiding consequent works such as site-directed mutagenesis and macromolecular docking.     

Improved secondary structure predictions for a nicotinic receptor subunit: incorporation of solvent accessibility and experimental data into a two-dimensional representation

Abstract A refined prediction of the nicotinic acetylcholine receptor (nAChR) subunits' secondary structure was computed with third-generation algorithms. The four selected programs, PHD, Predator, DSC, and NNSSP, based on different prediction approaches, were applied to each sequence of an alignment of nAChR and 5-HT3 receptor subunits, as well as a larger alignment with related subunit sequences from glycine and GABA receptors. A consensus prediction was computed for the nAChR subunits through a "winner takes all" method. By integrating the probabilities obtained with PHD, DSC, and NNSSP, this prediction was filtered in order to eliminate the singletons and to more precisely establish the structure limits (only 4% of the residues were modified). The final consensus secondary structure includes nine alpha-helices (24.2% of the residues, with an average length of 13.9 residues) and 17 beta-strands (22.5% of the residues, with an average length of 6.6 residues). The large extracellular domain is predicted to be mainly composed of beta-strands, with only two helices at the amino-terminal end. The transmembrane segments are predicted to be in a mixed alpha/beta topology (with a predominance of alpha-helices), with no known equivalent in the current protein database. The cytoplasmic domain is predicted to consist of two well-conserved amphipathic helices joined together by an unfolded stretch of variable length and sequence. In general, the segments predicted to occur in a periodic structure correspond to the more conserved regions, as defined by an analysis of sequence conservation per position performed on 152 superfamily members. The solvent accessibility of each residue was predicted from the multiple alignments with PHDacc. Each segment with more than three exposed residues was assumed to be external to the core protein. Overall, these data constitute an envelope of structural constraints. In a subsequent step, experimental data relative to the extracellular portion of the complete receptor were incorporated into the model. This led to a proposed two-dimensional representation of the secondary structure in which the peptide chain of the extracellular domain winds alternatively between the two interfaces of the subunit. Although this representation is not a tertiary structure and does not lead to predictions of specific beta-beta interaction, it should provide a basic framework for further mutagenesis investigations and for fold recognition (threading) searches.     

Evaluation of short-range interactions as secondary structure energies for protein fold and sequence recognition.

Short-range interactions for secondary structures of proteins are evaluated as potentials of mean force from the observed frequencies of secondary structures in known protein structures which are assumed to have an equilibrium distribution with the Boltzmann factor of secondary structure energies. A secondary conformation at each residue position in a protein is described by a tripeptide, including one nearest neighbor on each side. The secondary structure potentials are approximated as additive contributions from neighboring residues along the sequence. These are part of an empirical potential to provide a crude estimate of protein conformational energy at a residue level. Unlike previous works, interactions are decoupled into intrinsic potentials of residues, potentials of backbone-backbone interactions, and of side chain-backbone interactions. Also interactions are decoupled into one-body, two-body, and higher order interactions between peptide backbone and side chain and between backbones. These decouplings are essential to correctly evaluate the total secondary structure energy of a protein structure without overcounting interactions. Each interaction potential is evaluated separately by taking account of the correlation in the amino acid order of protein sequences. Interactions among side chains are neglected, because of the relatively limited number of protein structures. Proteins 1999;36:347-356. Published 1999 Wiley-Liss, Inc.     

Protein contact prediction using patterns of correlation

We describe a new method for using neural networks to predict residue contact pairs in a protein. The main inputs to the neural network are a set of 25 measures of correlated mutation between all pairs of residues in two "windows" of size 5 centered on the residues of interest. While the individual pair-wise correlations are a relatively weak predictor of contact, by training the network on windows of correlation the accuracy of prediction is significantly improved. The neural network is trained on a set of 100 proteins and then tested on a disjoint set of 1033 proteins of known structure. An average predictive accuracy of 21.7% is obtained taking the best L/2 predictions for each protein, where L is the sequence length. Taking the best L/10 predictions gives an average accuracy of 30.7%. The predictor is also tested on a set of 59 proteins from the CASP5 experiment. The accuracy is found to be relatively consistent across different sequence lengths, but to vary widely according to the secondary structure. Predictive accuracy is also found to improve by using multiple sequence alignments containing many sequences to calculate the correlations.