Binding interface prediction by combining protein–protein docking results

We developed a method called residue contact frequency (RCF), which uses the complex structures generated by the protein–protein docking algorithm ZDOCK to predict interface residues. Unlike interface prediction algorithms that are based on monomers alone, RCF is binding partner specific. We evaluated the performance of RCF using the area under the precision‐recall (PR) curve (AUC) on a large protein docking Benchmark. RCF (AUC = 0.44) performed as well as meta‐PPISP (AUC = 0.43), which is one of the best monomer‐based interface prediction methods. In addition, we test a support vector machine (SVM) to combine RCF with meta‐PPISP and another monomer‐based interface prediction algorithm Evolutionary Trace to further improve the performance. We found that the SVM that combined RCF and meta‐PPISP achieved the best performance (AUC = 0.47). We used RCF to predict the binding interfaces of proteins that can bind to multiple partners and RCF was able to correctly predict interface residues that are unique for the respective binding partners. Furthermore, we found that residues that contributed greatly to binding affinity (hotspot residues) had significantly higher RCF than other residues. Proteins 2014; 82:57–66. © 2013 Wiley Periodicals, Inc.     

Reliability measures for membrane protein topology prediction algorithms

We have developed reliability scores for five widely used membrane protein topology prediction methods, and have applied them both on a test set of 92 bacterial plasma membrane proteins with experimentally determined topologies and on all predicted helix bundle membrane proteins in three fully sequenced genomes: Escherichia coli, Saccharomyces cerevisiae and Caenorhabditis elegans. We show that the reliability scores work well for the TMHMM and MEMSAT methods, and that they allow the probability that the predicted topology is correct to be estimated for any protein. We further show that the available test set is biased towards high-scoring proteins when compared to the genome-wide data sets, and provide estimates for the expected prediction accuracy of TMHMM across the three genomes. Finally, we show that the performance of TMHMM is considerably better when limited experimental information (such as the in/out location of a protein's C terminus) is available, and estimate that at least ten percentage points in overall accuracy in whole-genome predictions can be gained in this way.     

Great interactions: How binding incorrect partners can teach us about protein recognition and function

Protein–protein interactions play a key part in most biological processes and understanding their mechanism is a fundamental problem leading to numerous practical applications. The prediction of protein binding sites in particular is of paramount importance since proteins now represent a major class of therapeutic targets. Amongst others methods, docking simulations between two proteins known to interact can be a useful tool for the prediction of likely binding patches on a protein surface. From the analysis of the protein interfaces generated by a massive cross‐docking experiment using the 168 proteins of the Docking Benchmark 2.0, where all possible protein pairs, and not only experimental ones, have been docked together, we show that it is also possible to predict a protein's binding residues without having any prior knowledge regarding its potential interaction partners. Evaluating the performance of cross‐docking predictions using the area under the specificity‐sensitivity ROC curve (AUC) leads to an AUC value of 0.77 for the complete benchmark (compared to the 0.5 AUC value obtained for random predictions). Furthermore, a new clustering analysis performed on the binding patches that are scattered on the protein surface show that their distribution and growth will depend on the protein's functional group. Finally, in several cases, the binding‐site predictions resulting from the cross‐docking simulations will lead to the identification of an alternate interface, which corresponds to the interaction with a biomolecular partner that is not included in the original benchmark. Proteins 2016; 84:1408–1421. © 2016 The Authors Proteins: Structure, Function, and Bioinformatics Published by Wiley Periodicals, Inc.     

Protein models: The Grand Challenge of protein docking

Characterization of life processes at the molecular level requires structural details of protein–protein interactions (PPIs). The number of experimentally determined protein structures accounts only for a fraction of known proteins. This gap has to be bridged by modeling, typically using experimentally determined structures as templates to model related proteins. The fraction of experimentally determined PPI structures is even smaller than that for the individual proteins, due to a larger number of interactions than the number of individual proteins, and a greater difficulty of crystallizing protein–protein complexes. The approaches to structural modeling of PPI (docking) often have to rely on modeled structures of the interactors, especially in the case of large PPI networks. Structures of modeled proteins are typically less accurate than the ones determined by X‐ray crystallography or nuclear magnetic resonance. Thus the utility of approaches to dock these structures should be assessed by thorough benchmarking, specifically designed for protein models. To be credible, such benchmarking has to be based on carefully curated sets of structures with levels of distortion typical for modeled proteins. This article presents such a suite of models built for the benchmark set of the X‐ray structures from the Dockground resource ( http://dockground.bioinformatics.ku.edu ) by a combination of homology modeling and Nudged Elastic Band method. For each monomer, six models were generated with predefined C^α root mean square deviation from the native structure (1, 2, …, 6 Å). The sets and the accompanying data provide a comprehensive resource for the development of docking methodology for modeled proteins. Proteins 2014; 82:278–287. © 2013 Wiley Periodicals, Inc.     

Coupling dynamics and evolutionary information with structure to identify protein regulatory and functional binding sites

Binding sites in proteins can be either specifically functional binding sites (active sites) that bind specific substrates with high affinity or regulatory binding sites (allosteric sites), that modulate the activity of functional binding sites through effector molecules. Owing to their significance in determining protein function, the identification of protein functional and regulatory binding sites is widely acknowledged as an important biological problem. In this work, we present a novel binding site prediction method, Active and Regulatory site Prediction (AR-Pred), which supplements protein geometry, evolutionary, and physicochemical features with information about protein dynamics to predict putative active and allosteric site residues. As the intrinsic dynamics of globular proteins plays an essential role in controlling binding events, we find it to be an important feature for the identification of protein binding sites. We train and validate our predictive models on multiple balanced training and validation sets with random forest machine learning and obtain an ensemble of discrete models for each prediction type. Our models for active site prediction yield a median area under the curve (AUC) of 91% and Matthews correlation coefficient (MCC) of 0.68, whereas the less well-defined allosteric sites are predicted at a lower level with a median AUC of 80% and MCC of 0.48. When tested on an independent set of proteins, our models for active site prediction show comparable performance to two existing methods and gains compared to two others, while the allosteric site models show gains when tested against three existing prediction methods. AR-Pred is available as a free downloadable package at https://github.com/sambitmishra0628/AR-PRED_source .     

Balanced ROC analysis (BAROC) protocol for the evaluation of protein similarities

Identification of problematic protein classes (domain types, protein families) that are difficult to predict from sequence is a key issue in genome annotation. ROC (Receiver Operating Characteristic) analysis is routinely used for the evaluation of protein similarities, however its results – the area under curve (AUC) values – are differentially biased for the various protein classes that are highly different in size. We show the bias can be compensated for by adjusting the length of the top list in a class-dependent fashion, so that the number of negatives within the top list will be equal to (or proportional with) the size of the positive class. Using this balanced protocol the problematic classes can be identified by their AUC values, or by a scatter diagram in which the AUC values are plotted against positive/negative ratio of the top list. The use of likelihood-ratio scoring (Kaján et al, Bioinformatics, 22, 2865–2869, 2007) the bias caused by class imbalance can be further decreased.     

Exploring the effects of sparse restraints on protein structure prediction

One of the main barriers to accurate computational protein structure prediction is searching the vast space of protein conformations. Distance restraints or inter‐residue contacts have been used to reduce this search space, easing the discovery of the correct folded state. It has been suggested that about 1 contact for every 12 residues may be sufficient to predict structure at fold level accuracy. Here, we use coarse‐grained structure‐based models in conjunction with molecular dynamics simulations to examine this empirical prediction. We generate sparse contact maps for 15 proteins of varying sequence lengths and topologies and find that given perfect secondary‐structural information, a small fraction of the native contact map (5%‐10%) suffices to fold proteins to their correct native states. We also find that different sparse maps are not equivalent and we make several observations about the type of maps that are successful at such structure prediction. Long range contacts are found to encode more information than shorter range ones, especially for α and αβ‐proteins. However, this distinction reduces for β‐proteins. Choosing contacts that are a consensus from successful maps gives predictive sparse maps as does choosing contacts that are well spread out over the protein structure. Additionally, the folding of proteins can also be used to choose predictive sparse maps. Overall, we conclude that structure‐based models can be used to understand the efficacy of structure‐prediction restraints and could, in future, be tuned to include specific force‐field interactions, secondary structure errors and noise in the sparse maps.     

Prediction of apoptosis protein subcellular location using improved hybrid approach and pseudo-amino acid composition

Apoptosis proteins are very important for understanding the mechanism of programmed cell death. The apoptosis protein localization can provide valuable information about its molecular function. The prediction of localization of an apoptosis protein is a challenging task. In our previous work we proposed an increment of diversity (ID) method using protein sequence information for this prediction task. In this work, based on the concept of Chou's pseudo-amino acid composition [Chou, K.C., 2001. Prediction of protein cellular attributes using pseudo-amino acid composition. Proteins: Struct. Funct. Genet. (Erratum: Chou, K.C., 2001, vol. 44, 60) 43, 246-255, Chou, K.C., 2005. Using amphiphilic pseudo-amino acid composition to predict enzyme subfamily classes. Bioinformatics 21, 10-19], a different pseudo-amino acid composition by using the hydropathy distribution information is introduced. A novel ID_SVM algorithm combined ID with support vector machine (SVM) is proposed. This method is applied to three data sets (317 apoptosis proteins, 225 apoptosis proteins and 98 apoptosis proteins). The higher predictive success rates than the previous algorithms are obtained by the jackknife tests.     

A framework for incorporating functional interrelationships into protein function prediction algorithms

The functional annotation of proteins is one of the most important tasks in the post-genomic era. Although many computational approaches have been developed in recent years to predict protein function, most of these traditional algorithms do not take interrelationships among functional terms into account, such as different GO terms usually coannotate with some common proteins. In this study, we propose a new functional similarity measure in the form of Jaccard coefficient to quantify these interrelationships and also develop a framework for incorporating GO term similarity into protein function prediction process. The experimental results of cross-validation on S. cerevisiae and Homo sapiens data sets demonstrate that our method is able to improve the performance of protein function prediction. In addition, we find that small size terms associated with a few of proteins obtain more benefit than the large size ones when considering functional interrelationships. We also compare our similarity measure with other two widely used measures, and results indicate that when incorporated into function prediction algorithms, our proposed measure is more effective. Experiment results also illustrate that our algorithms outperform two previous competing algorithms, which also take functional interrelationships into account, in prediction accuracy. Finally, we show that our method is robust to annotations in the database which are not complete at present. These results give new insights about the importance of functional interrelationships in protein function prediction.     

Detection of protein complexes using a protein ranking algorithm

Detecting protein complexes from protein‐protein interaction (PPI) network is becoming a difficult challenge in computational biology. There is ample evidence that many disease mechanisms involve protein complexes, and being able to predict these complexes is important to the characterization of the relevant disease for diagnostic and treatment purposes. This article introduces a novel method for detecting protein complexes from PPI by using a protein ranking algorithm (ProRank). ProRank quantifies the importance of each protein based on the interaction structure and the evolutionarily relationships between proteins in the network. A novel way of identifying essential proteins which are known for their critical role in mediating cellular processes and constructing protein complexes is proposed and analyzed. We evaluate the performance of ProRank using two PPI networks on two reference sets of protein complexes created from Munich Information Center for Protein Sequence, containing 81 and 162 known complexes, respectively. We compare the performance of ProRank to some of the well known protein complex prediction methods (ClusterONE, CMC, CFinder, MCL, MCode and Core) in terms of precision and recall. We show that ProRank predicts more complexes correctly at a competitive level of precision and recall. The level of the accuracy achieved using ProRank in comparison to other recent methods for detecting protein complexes is a strong argument in favor of the proposed method. Proteins 2012;. © 2012 Wiley Periodicals, Inc.     

Computational analysis of human protein interaction networks

Large amounts of human protein interaction data have been produced by experiments and prediction methods. However, the experimental coverage of the human interactome is still low in contrast to predicted data. To gain insight into the value of publicly available human protein network data, we compared predicted datasets, high-throughput results from yeast two-hybrid screens, and literature-curated protein-protein interactions. This evaluation is not only important for further methodological improvements, but also for increasing the confidence in functional hypotheses derived from predictions. Therefore, we assessed the quality and the potential bias of the different datasets using functional similarity based on the Gene Ontology, structural iPfam domain-domain interactions, likelihood ratios, and topological network parameters. This analysis revealed major differences between predicted datasets, but some of them also scored at least as high as the experimental ones regarding multiple quality measures. Therefore, since only small pair wise overlap between most datasets is observed, they may be combined to enlarge the available human interactome data. For this purpose, we additionally studied the influence of protein length on data quality and the number of disease proteins covered by each dataset. We could further demonstrate that protein interactions predicted by more than one method achieve an elevated reliability.     

A practical overview of protein disorder prediction methods

In the past few years there has been a growing awareness that a large number of proteins contain long disordered (unstructured) regions that often play a functional role. However, these disordered regions are still poorly detected. Recognition of disordered regions in a protein is important for two main reasons: reducing bias in sequence similarity analysis by avoiding alignment of disordered regions against ordered ones, and helping to delineate boundaries of protein domains to guide structural and functional studies. As none of the available method for disorder prediction can be taken as fully reliable on its own, we present an overview of the methods currently employed highlighting their advantages and drawbacks. We show a few practical examples of how they can be combined to avoid pitfalls and to achieve more reliable predictions.     

Hidden partners: Using cross‐docking calculations to predict binding sites for proteins with multiple interactions

Protein‐protein interactions control a large range of biological processes and their identification is essential to understand the underlying biological mechanisms. To complement experimental approaches, in silico methods are available to investigate protein‐protein interactions. Cross‐docking methods, in particular, can be used to predict protein binding sites. However, proteins can interact with numerous partners and can present multiple binding sites on their surface, which may alter the binding site prediction quality. We evaluate the binding site predictions obtained using complete cross‐docking simulations of 358 proteins with 2 different scoring schemes accounting for multiple binding sites. Despite overall good binding site prediction performances, 68 cases were still associated with very low prediction quality, presenting individual area under the specificity‐sensitivity ROC curve (AUC) values below the random AUC threshold of 0.5, since cross‐docking calculations can lead to the identification of alternate protein binding sites (that are different from the reference experimental sites). For the large majority of these proteins, we show that the predicted alternate binding sites correspond to interaction sites with hidden partners, that is, partners not included in the original cross‐docking dataset. Among those new partners, we find proteins, but also nucleic acid molecules. Finally, for proteins with multiple binding sites on their surface, we investigated the structural determinants associated with the binding sites the most targeted by the docking partners.     

A Simple Method for Predicting Transmembrane Proteins Based on Wavelet Transform

The increasing protein sequences from the genome project require theoretical methods to predict transmembrane helical segments (TMHs). So far, several prediction methods have been reported, but there are some deficiencies in prediction accuracy and adaptability in these methods. In this paper, a method based on discrete wavelet transform (DWT) has been developed to predict the number and location of TMHs in membrane proteins. PDB coded as 1KQG is chosen as an example to describe the prediction process by this method. 80 proteins with known 3D structure from Mptopo database are chosen at random as data sets (including 325 TMHs) and 80 sequences are divided into 13 groups according to their function and type. TMHs prediction is carried out for each group of membrane protein sequences and obtain satisfactory result. To verify the feasibility of this method, 80 membrane protein sequences are treated as test sets, 308 TMHs can be predicted and the prediction accuracy is 96.3%. Compared with the main prediction results of seven popular prediction methods, the obtained results indicate that the proposed method in this paper has higher prediction accuracy.     

PIER: protein interface recognition for structural proteomics

Recent advances in structural proteomics call for development of fast and reliable automatic methods for prediction of functional surfaces of proteins with known three-dimensional structure, including binding sites for known and unknown protein partners as well as oligomerization interfaces. Despite significant progress the problem is still far from being solved. Most existing methods rely, at least partially, on evolutionary information from multiple sequence alignments projected on protein surface. The common drawback of such methods is their limited applicability to the proteins with a sparse set of sequential homologs, as well as inability to detect interfaces in evolutionary variable regions. In this study, the authors developed an improved method for predicting interfaces from a single protein structure, which is based on local statistical properties of the protein surface derived at the level of atomic groups. The proposed Protein IntErface Recognition (PIER) method achieved the overall precision of 60% at the recall threshold of 50% at the residue level on a diverse benchmark of 490 homodimeric, 62 heterodimeric, and 196 transient interfaces (compared with 25% precision at 50% recall expected from random residue function assignment). For 70% of proteins in the benchmark, the binding patch residues were successfully detected with precision exceeding 50% at 50% recall. The calculation only took seconds for an average 300-residue protein. The authors demonstrated that adding the evolutionary conservation signal only marginally influenced the overall prediction performance on the benchmark; moreover, for certain classes of proteins, using this signal actually resulted in a deteriorated prediction. Thorough benchmarking using other datasets from literature showed that PIER yielded improved performance as compared with several alignment-free or alignment-dependent predictions. The accuracy, efficiency, and dependence on structure alone make PIER a suitable tool for automated high-throughput annotation of protein structures emerging from structural proteomics projects.     

Interactome mapping for analysis of complex phenotypes: insights from benchmarking binary interaction assays

Protein interactions mediate essentially all biological processes and analysis of protein-protein interactions using both large-scale and small-scale approaches has contributed fundamental insights to the understanding of biological systems. In recent years, interactome network maps have emerged as an important tool for analyzing and interpreting genetic data of complex phenotypes. Complementary experimental approaches to test for binary, direct interactions, and for membership in protein complexes are used to explore the interactome. The two approaches are not redundant but yield orthogonal perspectives onto the complex network of physical interactions by which proteins mediate biological processes. In recent years, several publications have demonstrated that interactions from high-throughput experiments can be equally reliable as the high quality subset of interactions identified in small-scale studies. Critical for this insight was the introduction of standardized experimental benchmarking of interaction and validation assays using reference sets. The data obtained in these benchmarking experiments have resulted in greater appreciation of the limitations and the complementary strengths of different assays. Moreover, benchmarking is a central element of a conceptual framework to estimate interactome sizes and thereby measure progress toward near complete network maps. These estimates have revealed that current large-scale data sets, although often of high quality, cover only a small fraction of a given interactome. Here, I review the findings of assay benchmarking and discuss implications for quality control, and for strategies toward obtaining a near-complete map of the interactome of an organism.     

Accuracy of protein flexibility predictions

Protein structural flexibility is important for catalysis, binding, and allostery. Flexibility has been predicted from amino acid sequence with a sliding window averaging technique and applied primarily to epitope search. New prediction parameters were derived from 92 refined protein structures in an unbiased selection of the Protein Data Bank by developing further the method of Karplus and Schulz (Naturwissenschaften 72:212–213, 1985). The accuracy of four flexibility prediction techniques was studied by comparing atomic temperature factors of known three-dimensional protein structures to predictions by using correlation coefficients. The size of the prediction window was optimized for each method. Predictions made with our new parameters, using an optimized window size of 9 residues in the prediction window, were giving the best results. The difference from another previously used technique was small, whereas two other methods were much poorer. Applicability of the predictions was also tested by searching for known epitopes from amino acid sequences. The best techniques predicted correctly 20 of 31 continuous epitopes in seven proteins. Flexibility parameters have previously been used for calculating protein average flexibility indices which are inversely correlated to protein stability. Indices with the new parameters showed better correlation to protein stability than those used previously; furthermore they had relationship even when the old parameters failed. © 1994 Wiley-Liss, Inc.     

Prediction of protein relative solvent accessibility with a two-stage SVM approach

Information on relative solvent accessibility (RSA) of amino acid residues in proteins provides valuable clues to the prediction of protein structure and function. A two-stage approach with support vector machines (SVMs) is proposed, where an SVM predictor is introduced to the output of the single-stage SVM approach to take into account the contextual relationships among solvent accessibilities for the prediction. By using the position-specific scoring matrices (PSSMs) generated by PSI-BLAST, the two-stage SVM approach achieves accuracies up to 90.4% and 90.2% on the Manesh data set of 215 protein structures and the RS126 data set of 126 nonhomologous globular proteins, respectively, which are better than the highest published scores on both data sets to date. A Web server for protein RSA prediction using a two-stage SVM method has been developed and is available (http://birc.ntu.edu.sg/~pas0186457/rsa.html).     

Using physical features of protein core packing to distinguish real proteins from decoys

The ability to consistently distinguish real protein structures from computationally generated model decoys is not yet a solved problem. One route to distinguish real protein structures from decoys is to delineate the important physical features that specify a real protein. For example, it has long been appreciated that the hydrophobic cores of proteins contribute significantly to their stability. We used two sources to obtain datasets of decoys to compare with real protein structures: submissions to the biennial Critical Assessment of protein Structure Prediction competition, in which researchers attempt to predict the structure of a protein only knowing its amino acid sequence, and also decoys generated by 3DRobot, which have user‐specified global root‐mean‐squared deviations from experimentally determined structures. Our analysis revealed that both sets of decoys possess cores that do not recapitulate the key features that define real protein cores. In particular, the model structures appear more densely packed (because of energetically unfavorable atomic overlaps), contain too few residues in the core, and have improper distributions of hydrophobic residues throughout the structure. Based on these observations, we developed a feed‐forward neural network, which incorporates key physical features of protein cores, to predict how well a computational model recapitulates the real protein structure without knowledge of the structure of the target sequence. By identifying the important features of protein structure, our method is able to rank decoy structures with similar accuracy to that obtained by state‐of‐the‐art methods that incorporate many additional features. The small number of physical features makes our model interpretable, emphasizing the importance of protein packing and hydrophobicity in protein structure prediction.