Data-driven docking: HADDOCK's adventures in CAPRI

Shape-Restrained Modeling of Protein-Small-Molecule Complexes with High Ambiguity Driven DOCKing

Small-molecule docking remains one of the most valuable computational techniques for the structure prediction of protein-small-molecule complexes. It allows us to study the interactions between compounds and the protein receptors they target at atomic detail in a timely and efficient manner. Here, we present a new protocol in HADDOCK (High Ambiguity Driven DOCKing), our integrative modeling platform, which incorporates homology information for both receptor and compounds. It makes use of HADDOCK's unique ability to integrate information in the simulation to drive it toward conformations, which agree with the provided data. The focal point is the use of shape restraints derived from homologous compounds bound to the target receptors. We have developed two protocols: in the first, the shape is composed of dummy atom beads based on the position of the heavy atoms of the homologous template compound, whereas in the second, the shape is additionally annotated with pharmacophore data for some or all beads. For both protocols, ambiguous distance restraints are subsequently defined between those beads and the heavy atoms of the ligand to be docked. We have benchmarked the performance of these protocols with a fully unbound version of the widely used DUD-E (Database of Useful Decoys-Enhanced) dataset. In this unbound docking scenario, our template/shape-based docking protocol reaches an overall success rate of 81% when a reliable template can be identified (which was the case for 99 out of 102 complexes in the DUD-E dataset), which is close to the best results reported for bound docking on the DUD-E dataset.

COVID-19 spike-host cell receptor GRP78 binding site prediction

OBJECTIVES

Understanding the novel coronavirus (COVID-19) mode of host cell recognition may help to fight the disease and save lives. The spike protein of coronaviruses is the main driving force for host cell recognition.

METHODS

In this study, the COVID-19 spike binding site to the cell-surface receptor (Glucose Regulated Protein 78 (GRP78)) is predicted using combined molecular modeling docking and structural bioinformatics. The COVID-19 spike protein is modeled using its counterpart, the SARS spike.

RESULTS

Sequence and structural alignments show that four regions, in addition to its cyclic nature have sequence and physicochemical similarities to the cyclic Pep42. Protein-protein docking was performed to test the four regions of the spike that fit tightly in the GRP78 Substrate Binding Domain β (SBDβ). The docking pose revealed the involvement of the SBDβ of GRP78 and the receptor-binding domain of the coronavirus spike protein in recognition of the host cell receptor.

CONCLUSIONS

We reveal that the binding is more favorable between regions III (C391-C525) and IV (C480-C488) of the spike protein model and GRP78. Region IV is the main driving force for GRP78 binding with the predicted binding affinity of -9.8 kcal/mol. These nine residues can be used to develop therapeutics specific against COVID-19.

Prediction of spacer-α6 complex: a novel insight into binding of ADAMTS13 with A2 domain of von Willebrand factor under forces

Force-regulated cleavage of A2 domain of von Willebrand factor (vWF) by ADAMTS13 is a key event in preventing thrombotic thrombocytopenic purpura (TTP). Recognition and cleavage depend on cooperative and modular contacts between several ADAMTS13 subdomains and discrete segments of vWF A2 domain. Spacer domain of ADAMTS13 contains an important exosite interacting with α6 helix of unfold A2 domain, but it remains unclear whether stretching of α6 regulates binding to spacer. To understand the molecular mechanism underlying the interactions between spacer and α6 under stretching, we successfully predicted spacer-α6 complex by a novel computer strategy combined the steered molecular dynamics (SMD) and flexible docking techniques. This strategy included three steps: (1) constant-velocity SMD simulation of α6; (2) zero-velocity SMD simulations of α6, and (3) flexible dockings of α6 to spacer. In our spacer-α6 complex model, 13 key residues, six in α6 and seven in spacer, were identified. Our data demonstrated a biphasic extension-regulated binding of α6 to spacer. The binding strength of the complex increased with α6 extension until it reaches its optimum of 0.25 nm, and then decreased as α6 extension further increased, meaning that spacer is in favor to binding with a partially extended α6, which may contribute to the optimal contact and proteolysis. Changes of interface area and intermolecular salt bridge may serve as the molecular basis for this characteristic. These findings provide a novel insight into mechano-chemical regulation on interaction between ADAMTS13 and vWF A2 domain under forces.

Improving protein-protein interaction prediction using evolutionary information from low-quality MSAs

Evolutionary information stored in multiple sequence alignments (MSAs) has been used to identify the interaction interface of protein complexes, by measuring either co-conservation or co-mutation of amino acid residues across the interface. Recently, maximum entropy related correlated mutation measures (CMMs) such as direct information, decoupling direct from indirect interactions, have been developed to identify residue pairs interacting across the protein complex interface. These studies have focussed on carefully selected protein complexes with large, good-quality MSAs. In this work, we study protein complexes with a more typical MSA consisting of fewer than 400 sequences, using a set of 79 intramolecular protein complexes. Using a maximum entropy based CMM at the residue level, we develop an interface level CMM score to be used in re-ranking docking decoys. We demonstrate that our interface level CMM score compares favourably to the complementarity trace score, an evolutionary information-based score measuring co-conservation, when combined with the number of interface residues, a knowledge-based potential and the variability score of individual amino acid sites. We also demonstrate, that, since co-mutation and co-complementarity in the MSA contain orthogonal information, the best prediction performance using evolutionary information can be achieved by combining the co-mutation information of the CMM with co-conservation information of a complementarity trace score, predicting a near-native structure as the top prediction for 41% of the dataset. The method presented is not restricted to small MSAs, and will likely improve interface prediction also for complexes with large and good-quality MSAs.

Sense and simplicity in HADDOCK scoring: Lessons from CASP-CAPRI round 1

Our information-driven docking approach HADDOCK is a consistent top predictor and scorer since the start of its participation in the CAPRI community-wide experiment. This sustained performance is due, in part, to its ability to integrate experimental data and/or bioinformatics information into the modelling process, and also to the overall robustness of the scoring function used to assess and rank the predictions. In the CASP-CAPRI Round 1 scoring experiment we successfully selected acceptable/medium quality models for 18/14 of the 25 targets - a top-ranking performance among all scorers. Considering that for only 20 targets acceptable models were generated by the community, our effective success rate reaches as high as 90% (18/20). This was achieved using the standard HADDOCK scoring function, which, thirteen years after its original publication, still consists of a simple linear combination of intermolecular van der Waals and Coulomb electrostatics energies and an empirically derived desolvation energy term. Despite its simplicity, this scoring function makes sense from a physico-chemical perspective, encoding key aspects of biomolecular recognition. In addition to its success in the scoring experiment, the HADDOCK server takes the first place in the server prediction category, with 16 successful predictions. Much like our scoring protocol, because of the limited time per target, the predictions relied mainly on either an ab initio center-of-mass and symmetry restrained protocol, or on a template-based approach whenever applicable. These results underline the success of our simple but sensible prediction and scoring scheme. Proteins 2017; 85:417-423. © 2016 Wiley Periodicals, Inc.

dMM-PBSA: A New HADDOCK Scoring Function for Protein-Peptide Docking

Molecular-docking programs coupled with suitable scoring functions are now established and very useful tools enabling computational chemists to rapidly screen large chemical databases and thereby to identify promising candidate compounds for further experimental processing. In a broader scenario, predicting binding affinity is one of the most critical and challenging components of computer-aided structure-based drug design. The development of a molecular docking scoring function which in principle could combine both features, namely ranking putative poses and predicting complex affinity, would be of paramount importance. Here, we systematically investigated the performance of the MM-PBSA approach, using two different Poisson-Boltzmann solvers (APBS and DelPhi), in the currently rising field of protein-peptide interactions (PPIs), identifying the correct binding conformations of 19 different protein-peptide complexes and predicting their binding free energies. First, we scored the decoy structures from HADDOCK calculation via the MM-PBSA approach in order to assess the capability of retrieving near-native poses in the best-scoring clusters and of evaluating the corresponding free energies of binding. MM-PBSA behaves well in finding the poses corresponding to the lowest binding free energy, however the built-in HADDOCK score shows a better performance. In order to improve the MM-PBSA-based scoring function, we dampened the MM-PBSA solvation and coulombic terms by 0.2, as proposed in the HADDOCK score and LIE approaches. The new dampened MM-PBSA (dMM-PBSA) outperforms the original MM-PBSA and ranks the decoys structures as the HADDOCK score does. Second, we found a good correlation between the dMM-PBSA and HADDOCK scores for the near-native clusters of each system and the experimental binding energies, respectively. Therefore, we propose a new scoring function, dMM-PBSA, to be used together with the built-in HADDOCK score in the context of protein-peptide docking simulations.

Performance of ZDOCK in CAPRI rounds 20-26

We report the performance of our approaches for protein-protein docking and interface analysis in CAPRI rounds 20-26. At the core of our pipeline was the ZDOCK program for rigid-body protein-protein docking. We then reranked the ZDOCK predictions using the ZRANK or IRAD scoring functions, pruned and analyzed energy landscapes using clustering, and analyzed the docking results using our interface prediction approach RCF. When possible, we used biological information from the literature to apply constraints to the search space during or after the ZDOCK runs. For approximately half of the standard docking challenges we made at least one prediction that was acceptable or better. For the scoring challenges we made acceptable or better predictions for all but one target. This indicates that our scoring functions are generally able to select the correct binding mode.

Robust Antibody-Antigen Complexes Prediction Generated by Combining Sequence Analyses, Mutagenesis, In Vitro Evolution, X-ray Crystallography and In Silico Docking

Hu 15C1 is a potent anti-human Toll-like receptor 4 (TLR4) neutralizing antibody. To better understand the molecular basis of its biological activity, we used a multidisciplinary approach to generate an accurate model of the Hu 15C1-TLR4 complex. By combining site-directed mutagenesis, in vitro antibody evolution, affinity measurements and X-ray crystallography of Fab fragments, we identified key interactions across the Hu 15C1-TLR4 interface. These contact points were used as restraints to predict the structure of the Fab region of Hu 15C1 bound to TLR4 using computational molecular docking. This model was further evaluated and validated by additional site-directed mutagenesis studies. The predicted structure of the Hu 15C1-TLR4 complex indicates that the antibody antagonizes the receptor dimerization necessary for its activation. This study exemplifies how iterative cycles of antibody engineering can facilitate the discovery of components of antibody-target interactions.

FoxO1 negatively regulates leptin-induced POMC transcription through its direct interaction with STAT3

The amino acid region Gly140–Leu160 (key residues Gln145, Arg147, Lys148, Arg153 and Arg154) of FoxO1 is identified as a STAT3 binding site, which is critical for FoxO1 to inhibit STAT3-mediated leptin-induced POMC activation.

Combined computational and experimental analysis of a complex of ribonuclease III and the regulatory macrodomain protein, YmdB

Ribonuclease III is a conserved bacterial endonuclease that cleaves double-stranded(ds) structures in diverse coding and noncoding RNAs. RNase III is subject to multiple levels of control that in turn confer global post-transcriptional regulation. The Escherichia coli macrodomain protein YmdB directly interacts with RNase III, and an increase in YmdB amount in vivo correlates with a reduction in RNase III activity. Here, a computational-based structural analysis was performed to identify atomic-level features of the YmdB-RNase III interaction. The docking of monomeric E. coli YmdB with a homology model of the E. coli RNase III homodimer yields a complex that exhibits an interaction of the conserved YmdB residue R40 with specific RNase III residues at the subunit interface. Surface Plasmon Resonance (SPR) analysis provided a K_D of 61 nM for the complex, corresponding to a binding free energy (ΔG) of −9.9 kcal/mol. YmdB R40 and RNase III D128 were identified by in silico alanine mutagenesis as thermodynamically important interacting partners. Consistent with the prediction, the YmdB R40A mutation causes a 16-fold increase in K_D (ΔΔG = +1.8 kcal/mol), as measured by SPR, and the D128A mutation in both RNase III subunits (D128A/D128′A) causes an 83-fold increase in K_D (ΔΔG = +2.7 kcal/mol). The greater effect of the D128A/D128′A mutation may reflect an altered RNase III secondary structure, as revealed by CD spectroscopy, which also may explain the significant reduction in catalytic activity in vitro. The features of the modeled complex relevant to potential RNase III regulatory mechanisms are discussed. Proteins 2015; 83:459–472. © 2014 The Authors. Proteins: Structure, Function, and Bioinformatics Published by Wiley Periodicals, Inc.

Defining the limits of homology modeling in information-driven protein docking

Information-driven docking is currently one of the most successful approaches to obtain structural models of protein interactions as demonstrated in the latest round of CAPRI. While various experimental and computational techniques can be used to retrieve information about the binding mode, the availability of three-dimensional structures of the interacting partners remains a limiting factor. Fortunately, the wealth of structural information gathered by large-scale initiatives allows for homology-based modeling of a significant fraction of the protein universe. Defining the limits of information-driven docking based on such homology models is therefore highly relevant. Here we show, using previous CAPRI targets, that out of a variety of measures, the global sequence identity between template and target is a simple but reliable predictor of the achievable quality of the docking models. This indicates that a well-defined overall fold is critical for the interaction. Furthermore, the quality of the data at our disposal to characterize the interaction plays a determinant role in the success of the docking. Given reliable interface information we can obtain acceptable predictions even at low global sequence identity. These results, which define the boundaries between trustworthy and unreliable predictions, should guide both experts and nonexperts in defining the limits of what is achievable by docking. This is highly relevant considering that the fraction of the interactome amenable for docking is only bound to grow as the number of experimentally solved structures increases.

Structural determinants at the interface of the ARC2 and leucine-rich repeat domains control the activation of the plant immune receptors Rx1 and Gpa2

Many plant and animal immune receptors have a modular nucleotide-binding-leucine-rich repeat (NB-LRR) architecture in which a nucleotide-binding switch domain, NB-ARC, is tethered to a LRR sensor domain. The cooperation between the switch and sensor domains, which regulates the activation of these proteins, is poorly understood. Here, we report structural determinants governing the interaction between the NB-ARC and LRR in the highly homologous plant immune receptors Gpa2 and Rx1, which recognize the potato cyst nematode Globodera pallida and Potato virus X, respectively. Systematic shuffling of polymorphic sites between Gpa2 and Rx1 showed that a minimal region in the ARC2 and N-terminal repeats of the LRR domain coordinate the activation state of the protein. We identified two closely spaced amino acid residues in this region of the ARC2 (positions 401 and 403) that distinguish between autoactivation and effector-triggered activation. Furthermore, a highly acidic loop region in the ARC2 domain and basic patches in the N-terminal end of the LRR domain were demonstrated to be required for the physical interaction between the ARC2 and LRR. The NB-ARC and LRR domains dissociate upon effector-dependent activation, and the complementary-charged regions are predicted to mediate a fast reassociation, enabling multiple rounds of activation. Finally, we present a mechanistic model showing how the ARC2, NB, and N-terminal half of the LRR form a clamp, which regulates the dissociation and reassociation of the switch and sensor domains in NB-LRR proteins.

Virtual interactomics of proteins from biochemical standpoint

Virtual interactomics represents a rapidly developing scientific area on the boundary line of bioinformatics and interactomics. Protein-related virtual interactomics then comprises instrumental tools for prediction, simulation, and networking of the majority of interactions important for structural and individual reproduction, differentiation, recognition, signaling, regulation, and metabolic pathways of cells and organisms. Here, we describe the main areas of virtual protein interactomics, that is, structurally based comparative analysis and prediction of functionally important interacting sites, mimotope-assisted and combined epitope prediction, molecular (protein) docking studies, and investigation of protein interaction networks. Detailed information about some interesting methodological approaches and online accessible programs or databases is displayed in our tables. Considerable part of the text deals with the searches for common conserved or functionally convergent protein regions and subgraphs of conserved interaction networks, new outstanding trends and clinically interesting results. In agreement with the presented data and relationships, virtual interactomic tools improve our scientific knowledge, help us to formulate working hypotheses, and they frequently also mediate variously important in silico simulations.

PI2PE: A Suite of Web Servers for Predictions Ranging From Protein Structure to Binding Kinetics

PI²PE (http://pipe.sc.fsu.edu) is a suite of four web servers for predicting a variety of folding- and binding-related properties of proteins. These include the solvent accessibility of amino acids upon protein folding, the amino acids forming the interfaces of protein-protein and protein-nucleic acid complexes, and the binding rate constants of these complexes. Three of the servers debuted in 2007, and have garnered ~2,500 unique users and finished over 30,000 jobs. The functionalities of these servers are now enhanced, and a new sever, for predicting the binding rate constants, is added. Together, these web servers form a pipeline from protein sequence to tertiary structure, then to quaternary structure, and finally to binding kinetics.

Structural models of protein-DNA complexes based on interface prediction and docking

Protein-DNA interactions are the physical basis of gene expression and DNA modification. Structural models that reveal these interactions are essential for their understanding. As only a limited number of structures for protein-DNA complexes have been determined by experimental methods, computation methods provide a potential way to fill the need. We have developed the DISPLAR method to predict DNA binding sites on proteins. Predicted binding sites have been used to assist the building of structural models by docking, either by guiding the docking or by selecting near-native candidates from the docked poses. Here we applied the DISPLAR method to predict the DNA binding sites for 20 DNA-binding proteins, which have had their DNA binding sites characterized by NMR chemical shift perturbation. For two of these proteins, the structures of their complexes with DNA have also been determined. With the help of the DISPLAR predictions, we built structural models for these two complexes. Evaluations of both the DNA binding sites for 20 proteins and the structural models of the two protein-DNA complexes against experimental results demonstrate the significant promise of our model-building approach.

Optimized grid-based protein-protein docking as a global search tool followed by incorporating experimentally derivable restraints

Unbound protein docking, or the computational prediction of the structure of a protein complex from the structures of its separated components, is of importance but still challenging. A practical approach toward reliable results for unbound docking is to incorporate experimentally derived information with computation. To this end, truly systematic search of the global docking space is desirable. The fast Fourier transform (FFT) docking is a systematic search method with high computational efficiency. However, by using FFT to perform unbound docking, possible conformational changes upon binding must be treated implicitly. To better accommodate the implicit treatment of conformational flexibility, we develop a rational approach to optimize "softened" parameters for FFT docking. In connection with the increased "softness" of the parameters in this global search step, we use a revised rule to select candidate models from the search results. For complexes designated as of low and medium difficulty for unbound docking, these adaptations of the original FTDOCK program lead to substantial improvements of the global search results. Finally, we show that models resulted from FFT-based global search can be further filtered with restraints derivable from nuclear magnetic resonance (NMR) chemical shift perturbation or mutagenesis experiments, leading to a small set of models that can be feasibly refined and evaluated using computationally more expensive methods and that still include high-ranking near-native conformations.

Prediction of protein-protein interaction sites using patch-based residue characterization

Identifying protein-protein interaction sites provides important clues to the function of a protein and is becoming increasingly relevant in topics such as systems biology and drug discovery. Using a patch-based model for residue characterization, we trained random forest classifiers for residue-based interface prediction, which was followed by a clustering procedure to produce patches for patch-based interface prediction. For residue-based interface prediction, our method achieves a specificity rate of 0.7 and a sensitivity rate of 0.78. For patch-based interface prediction, a success rate of 0.80 is achieved. Based on same datasets, we also compare it with several published methods. The results show that our method is a successful predictor for residue-based and patch-based interface prediction.

Structural model of ligand-G protein-coupled receptor (GPCR) complex based on experimental double mutant cycle data: MT7 snake toxin bound to dimeric hM1 muscarinic receptor

The snake toxin MT7 is a potent and specific allosteric modulator of the human M1 muscarinic receptor (hM1). We previously characterized by mutagenesis experiments the functional determinants of the MT7-hM1 receptor interaction (Fruchart-Gaillard, C., Mourier, G., Marquer, C., Stura, E., Birdsall, N. J., and Servent, D. (2008) Mol. Pharmacol. 74, 1554-1563) and more recently collected evidence indicating that MT7 may bind to a dimeric form of hM1 (Marquer, C., Fruchart-Gaillard, C., Mourier, G., Grandjean, O., Girard, E., le Maire, M., Brown, S., and Servent, D. (2010) Biol. Cell 102, 409-420). To structurally characterize the MT7-hM1 complex, we adopted a strategy combining double mutant cycle experiments and molecular modeling calculations. First, thirty-three ligand-receptor proximities were identified from the analysis of sixty-one double mutant binding affinities. Several toxin residues that are more than 25 Å apart still contact the same residues on the receptor. As a consequence, attempts to satisfy all the restraints by docking the toxin onto a single receptor failed. The toxin was then positioned onto two receptors during five independent flexible docking simulations. The different possible ligand and receptor extracellular loop conformations were described by performing simulations in explicit solvent. All the docking calculations converged to the same conformation of the MT7-hM1 dimer complex, satisfying the experimental restraints and in which (i) the toxin interacts with the extracellular side of the receptor, (ii) the tips of MT7 loops II and III contact one hM1 protomer, whereas the tip of loop I binds to the other protomer, and (iii) the hM1 dimeric interface involves the transmembrane helices TM6 and TM7. These results structurally support the high affinity and selectivity of the MT7-hM1 interaction and highlight the atypical mode of interaction of this allosteric ligand on its G protein-coupled receptor target.

Rapid identification of protein-protein interfaces for the construction of a complex model based on multiple unassigned signals by using time-sharing NMR measurements

Protein-protein interactions are necessary for various cellular processes, and therefore, information related to protein-protein interactions and structural information of complexes is invaluable. To identify protein-protein interfaces using NMR, resonance assignments are generally necessary to analyze the data; however, they are time consuming to collect, especially for large proteins. In this paper, we present a rapid, effective, and unbiased approach for the identification of a protein-protein interface without resonance assignments. This approach requires only a single set of 2D titration experiments of a single protein sample, labeled with a unique combination of an (15)N-labeled amino acid and several amino acids (13)C-labeled on specific atoms. To rapidly obtain high resolution data, we applied a new pulse sequence for time-shared NMR measurements that allowed simultaneous detection of a ω(1)-TROSY-type backbone (1)H-(15)N and aromatic (1)H-(13)C shift correlations together with single quantum methyl (1)H-(13)C shift correlations. We developed a structure-based computational approach, that uses our experimental data to search the protein surfaces in an unbiased manner to identify the residues involved in the protein-protein interface. Finally, we demonstrated that the obtained information of the molecular interface could be directly leveraged to support protein-protein docking studies. Such rapid construction of a complex model provides valuable information and enables more efficient biochemical characterization of a protein-protein complex, for instance, as the first step in structure-guided drug development.

Strengths and weaknesses of data-driven docking in critical assessment of prediction of interactions

The recent CAPRI rounds have introduced new docking challenges in the form of protein-RNA complexes, multiple alternative interfaces, and an unprecedented number of targets for which homology modeling was required. We present here the performance of HADDOCK and its web server in the CAPRI experiment and discuss the strengths and weaknesses of data-driven docking. HADDOCK was successful for 6 out of 9 complexes (6 out of 11 targets) and accurately predicted the individual interfaces for two more complexes. The HADDOCK server, which is the first allowing the simultaneous docking of generic multi-body complexes, was successful in 4 out of 7 complexes for which it participated. In the scoring experiment, we predicted the highest number of targets of any group. The main weakness of data-driven docking revealed from these last CAPRI results is its vulnerability for incorrect experimental data related to the interface or the stoichiometry of the complex. At the same time, the use of experimental and/or predicted information is also the strength of our approach as evidenced for those targets for which accurate experimental information was available (e.g., the 10 three-stars predictions for T40!). Even when the models show a wrong orientation, the individual interfaces are generally well predicted with an average coverage of 60% ± 26% over all targets. This makes data-driven docking particularly valuable in a biological context to guide experimental studies like, for example, targeted mutagenesis.

Selection of near-native poses in CAPRI rounds 13-19

In CAPRI rounds 13-19, we submitted models that are of acceptable or higher quality for 6 of the total of 13 targets. This success builds on our record in previous CAPRI rounds. The docking problem can be divided into two steps. In the first, translational/rotational and conformational space is searched to generate a pool of docked poses; the success of this search step is measured by whether near-native poses are included in the pool. In the second step, the pool is selected for near-native poses. In our previous assessment of CAPRI results, we suggested that the search problem is largely solved; a remaining problem is to select near-native poses. Our work in these new rounds of CAPRI was guided by this assessment. To solve the selection problem, we used an assortment of criteria on the interfaces of candidate poses. In one extreme, represented by T29, with very little known interface information, our criterion for top models was based on interface prediction. Poses in which the predicted interface residues occurred in interfaces were selected. Our model 1 for T29 was of medium quality. In the other extreme, represented by T40, with reliably known interface information, our selection was solely based on such information. Nine of the ten models submitted for T40 were of high (3 models), medium (4 models), and acceptable (2 models) quality. Our strategy of mixing predicted and known interface information appears to be widely applicable for the selection of near-native poses.

Accounting for large amplitude protein deformation during in silico macromolecular docking

Rapid progress of theoretical methods and computer calculation resources has turned in silico methods into a conceivable tool to predict the 3D structure of macromolecular assemblages, starting from the structure of their separate elements. Still, some classes of complexes represent a real challenge for macromolecular docking methods. In these complexes, protein parts like loops or domains undergo large amplitude deformations upon association, thus remodeling the surface accessible to the partner protein or DNA. We discuss the problems linked with managing such rearrangements in docking methods and we review strategies that are presently being explored, as well as their limitations and success.

Ranking Docked Models of Protein-Protein Complexes Using Predicted Partner-Specific Protein-Protein Interfaces: A Preliminary Study

Computational protein-protein docking is a valuable tool for determining the conformation of complexes formed by interacting proteins. Selecting near-native conformations from the large number of possible models generated by docking software presents a significant challenge in practice. We introduce a novel method for ranking docked conformations based on the degree of overlap between the interface residues of a docked conformation formed by a pair of proteins with the set of predicted interface residues between them. Our approach relies on a method, called PS-HomPPI, for reliably predicting protein-protein interface residues by taking into account information derived from both interacting proteins. PS-HomPPI infers the residues of a query protein that are likely to interact with a partner protein based on known interface residues of the homo-interologs of the query-partner protein pair, i.e., pairs of interacting proteins that are homologous to the query protein and partner protein. Our results on Docking Benchmark 3.0 show that the quality of the ranking of docked conformations using our method is consistently superior to that produced using ClusPro cluster-size-based and energy-based criteria for 61 out of the 64 docking complexes for which PS-HomPPI produces interface predictions. An implementation of our method for ranking docked models is freely available at: http://einstein.cs.iastate.edu/DockRank/.

Nucleotide docking: prediction of reactant state complexes for ribonuclease enzymes

Ribonuclease enzymes (RNases) play key roles in the maturation and metabolism of all RNA molecules. Computational simulations of the processes involved can help to elucidate the underlying enzymatic mechanism and is often employed in a synergistic approach together with biochemical experiments. Theoretical calculations require atomistic details regarding the starting geometries of the molecules involved, which, in the absence of crystallographic data, can only be achieved from computational docking studies. Fortunately, docking algorithms have improved tremendously in recent years, so that reliable structures of enzyme-ligand complexes can now be successfully obtained from computation. However, most docking programs are not particularly optimized for nucleotide docking. In order to assist our studies on the cleavage of RNA by the two most important ribonuclease enzymes, RNase A and RNase H, we evaluated four docking tools-MOE2009, Glide 5.5, QXP-Flo+0802, and Autodock 4.0-for their ability to simulate complexes between these enzymes and RNA oligomers. To validate our results, we analyzed the docking results with respect to the known key interactions between the protein and the nucleotide. In addition, we compared the predicted complexes with X-ray structures of the mutated enzyme as well as with structures obtained from previous calculations. In this manner, we were able to prepare the desired reaction state complex so that it could be used as the starting structure for further DFT/B3LYP QM/MM reaction mechanism studies.

Delineation of the complement receptor type 2-C3d complex by site-directed mutagenesis and molecular docking

The interactions between the complement receptor type 2 (CR2) and the C3 complement fragments C3d, C3dg, and iC3b are essential for the initiation of a normal immune response. A crystal-derived structure of the two N-terminal short consensus repeat (SCR1-2) domains of CR2 in complex with C3d has previously been elucidated. However, a number of biochemical and biophysical studies targeting both CR2 and C3d appear to be in conflict with these structural data. Previous mutagenesis and heteronuclear NMR spectroscopy studies directed toward the C3d-binding site on CR2 have indicated that the CR2-C3d cocrystal structure may represent an encounter/intermediate or nonphysiological complex. With regard to the CR2-binding site on C3d, mutagenesis studies by Isenman and coworkers [Isenman, D. E., Leung, E., Mackay, J. D., Bagby, S. & van den Elsen, J. M. H. (2010). Mutational analyses reveal that the staphylococcal immune evasion molecule Sbi and complement receptor 2 (CR2) share overlapping contact residues on C3d: Implications for the controversy regarding the CR2/C3d cocrystal structure. J. Immunol. 184, 1946-1955] have implicated an electronegative "concave" surface on C3d in the binding process. This surface is discrete from the CR2-C3d interface identified in the crystal structure. We generated a total of 18 mutations targeting the two (X-ray crystallographic- and mutagenesis-based) proposed CR2 SCR1-2 binding sites on C3d. Using ELISA analyses, we were able to assess binding of mutant forms of C3d to CR2. Mutations directed toward the concave surface of C3d result in substantially compromised CR2 binding. By contrast, targeting the CR2-C3d interface identified in the cocrystal structure and the surrounding area results in significantly lower levels of disruption in binding. Molecular modeling approaches used to investigate disparities between the biochemical data and the X-ray structure of the CR2-C3d cocrystal result in highest-scoring solutions in which CR2 SCR1-2 is docked within the concave surface of C3d.

Macromolecular docking restrained by a small angle X-ray scattering profile

While many structures of single protein components are becoming available, structural characterization of their complexes remains challenging. Methods for modeling assembly structures from individual components frequently suffer from large errors, due to protein flexibility and inaccurate scoring functions. However, when additional information is available, it may be possible to reduce the errors and compute near-native complex structures. One such type of information is a small angle X-ray scattering (SAXS) profile that can be collected in a high-throughput fashion from a small amount of sample in solution. Here, we present an efficient method for protein-protein docking with a SAXS profile (FoXSDock): generation of complex models by rigid global docking with PatchDock, filtering of the models based on the SAXS profile, clustering of the models, and refining the interface by flexible docking with FireDock. FoXSDock is benchmarked on 124 protein complexes with simulated SAXS profiles, as well as on 6 complexes with experimentally determined SAXS profiles. When induced fit is less than 1.5Å interface C(α) RMSD and the fraction residues of missing from the component structures is less than 3%, FoXSDock can find a model close to the native structure within the top 10 predictions in 77% of the cases; in comparison, docking alone succeeds in only 34% of the cases. Thus, the integrative approach significantly improves on molecular docking alone. The improvement arises from an increased resolution of rigid docking sampling and more accurate scoring.

Building macromolecular assemblies by information-driven docking: introducing the HADDOCK multibody docking server

Over the last years, large scale proteomics studies have generated a wealth of information of biomolecular complexes. Adding the structural dimension to the resulting interactomes represents a major challenge that classical structural experimental methods alone will have difficulties to confront. To meet this challenge, complementary modeling techniques such as docking are thus needed. Among the current docking methods, HADDOCK (High Ambiguity-Driven DOCKing) distinguishes itself from others by the use of experimental and/or bioinformatics data to drive the modeling process and has shown a strong performance in the critical assessment of prediction of interactions (CAPRI), a blind experiment for the prediction of interactions. Although most docking programs are limited to binary complexes, HADDOCK can deal with multiple molecules (up to six), a capability that will be required to build large macromolecular assemblies. We present here a novel web interface of HADDOCK that allows the user to dock up to six biomolecules simultaneously. This interface allows the inclusion of a large variety of both experimental and/or bioinformatics data and supports several types of cyclic and dihedral symmetries in the docking of multibody assemblies. The server was tested on a benchmark of six cases, containing five symmetric homo-oligomeric protein complexes and one symmetric protein-DNA complex. Our results reveal that, in the presence of either bioinformatics and/or experimental data, HADDOCK shows an excellent performance: in all cases, HADDOCK was able to generate good to high quality solutions and ranked them at the top, demonstrating its ability to model symmetric multicomponent assemblies. Docking methods can thus play an important role in adding the structural dimension to interactomes. However, although the current docking methodologies were successful for a vast range of cases, considering the variety and complexity of macromolecular assemblies, inclusion of some kind of experimental information (e.g. from mass spectrometry, nuclear magnetic resonance, cryoelectron microscopy, etc.) will remain highly desirable to obtain reliable results.

Rapid structural characterization of human antibody-antigen complexes through experimentally validated computational docking

If we understand the structural rules governing antibody (Ab)-antigen (Ag) interactions in a given virus, then we have the molecular basis to attempt to design and synthesize new epitopes to be used as vaccines or optimize the antibodies themselves for passive immunization. Comparing the binding of several different antibodies to related Ags should also further our understanding of general principles of recognition. To obtain and compare the three-dimensional structure of a large number of different complexes, however, we need a faster method than traditional experimental techniques. While biocomputational docking is fast, its results might not be accurate. Combining experimental validation with computational prediction may be a solution. As a proof of concept, here we isolated a monoclonal Ab from the blood of a human donor recovered from dengue virus infection, characterized its immunological properties, and identified its epitope on domain III of dengue virus E protein through simple and rapid NMR chemical shift mapping experiments. We then obtained the three-dimensional structure of the Ab/Ag complex by computational docking, using the NMR data to drive and validate the results. In an attempt to represent the multiple conformations available to flexible Ab loops, we docked several different starting models and present the result as an ensemble of models equally agreeing with the experimental data. The Ab was shown to bind a region accessible only in part on the viral surface, explaining why it cannot effectively neutralize the virus.

Analysis and prediction of protein quaternary structure

The quaternary structure (QS) of a protein is determined by measuring its molecular weight in solution. The data have to be extracted from the literature, and they may be missing even for proteins that have a crystal structure reported in the Protein Data Bank (PDB). The PDB and other databases derived from it report QS information that either was obtained from the depositors or is based on an analysis of the contacts between polypeptide chains in the crystal, and this frequently differs from the QS determined in solution.The QS of a protein can be predicted from its sequence using either homology or threading methods. However, a majority of the proteins with less than 30% sequence identity have different QSs. A model of the QS can also be derived by docking the subunits when their 3D structure is independently known, but the model is likely to be incorrect if large conformation changes take place when the oligomer assembles.

Development of a Novel Bioinformatics Tool for In Silico Validation of Protein Interactions

Protein interactions are crucial in most biological processes. Several in silico methods have been recently developed to predict them. This paper describes a bioinformatics method that combines sequence similarity and structural information to support experimental studies on protein interactions. Given a target protein, the approach selects the most likely interactors among the candidates revealed by experimental techniques, but not yet in vivo validated. The sequence and the structural information of the in vivo confirmed proteins and complexes are exploited to evaluate the candidate interactors. Finally, a score is calculated to suggest the most likely interactors of the target protein. As an example, we searched for GRB2 interactors. We ranked a set of 46 candidate interactors by the presented method. These candidates were then reduced to 21, through a score threshold chosen by means of a cross-validation strategy. Among them, the isoform 1 of MAPK14 was in silico confirmed as a GRB2 interactor. Finally, given a set of already confirmed interactors of GRB2, the accuracy and the precision of the approach were 75% and 86%, respectively. In conclusion, the proposed method can be conveniently exploited to select the proteins to be experimentally investigated within a set of potential interactors.

Allosteric communication between protomers of dopamine class A GPCR dimers modulates activation

A major obstacle to understanding the functional importance of dimerization between Class A G protein-coupled receptors (GPCRs) has been the methodological limitation in achieving control of the identity of the components comprising the signaling unit. We have developed a functional complementation assay that enables such control and illustrate it for the human dopamine D2 receptor. The minimal signaling unit, two receptors and a single G protein, is maximally activated by agonist binding to a single protomer, which suggests an asymmetrical activated dimer. Inverse agonist binding to the second protomer enhances signaling, whereas agonist binding to the second protomer blunts signaling. Ligand-independent constitutive activation of the second protomer also inhibits signaling. Thus, GPCR dimer function can be modulated by the activity state of the second protomer, which for a heterodimer may be altered in pathological states. Our novel methodology also makes possible the characterization of signaling from a defined heterodimer unit.

Rational redesign of neutral endopeptidase binding to merlin and moesin proteins

Neutral endopeptidase (NEP) is a 90- to 110-kDa cell-surface peptidase that is normally expressed by numerous tissues but whose expression is lost or reduced in a variety of malignancies. The anti-tumorigenic function of NEP is mediated not only by its catalytic activity but also through direct protein-protein interactions of its cytosolic region with several binding partners, including Lyn kinase, PTEN, and ezrin/radixin/moesin (ERM) proteins. We have previously shown that mutation of the K(19)K(20)K(21) basic cluster in NEPs' cytosolic region to residues QNI disrupts binding to the ERM proteins. Here we show that the ERM-related protein merlin (NF2) does not bind NEP or its cytosolic region. Using experimental data, threading, and sequence analysis, we predicted the involvement of moesin residues E(159)Q(160) in binding to the NEP cytosolic domain. Mutation of these residues to NL (to mimic the corresponding N(159)L(160) residues in the nonbinder merlin) disrupted moesin binding to NEP. Mutation of residues N(159)L(160)Y(161)K(162)M(163) in merlin to the corresponding moesin residues resulted in NEP binding to merlin. This engineered NEP peptide-merlin interaction was diminished by the QNI mutation in NEP, supporting the role of the NEP basic cluster in binding. We thus identified the region of interaction between NEP and moesin, and engineered merlin into a NEP-binding protein. These data form the basis for further exploration of the details of NEP-ERM binding and function.

Use of residual dipolar couplings in structural analysis of protein-ligand complexes by solution NMR spectroscopy

Investigation of structure-function relationships in protein complexes, specifically protein-ligand interactions, carry great significance in elucidating the structural and mechanistic bases of molecular recognition events and their role in regulating cell processes. Nuclear magnetic resonance (NMR) spectroscopy is one of the leading structural and analytical techniques in in-depth studies of protein-ligand interactions. Recent advances in NMR methodology such as transverse relaxation-optimized spectroscopy (TROSY) and residual dipolar couplings (RDCs) measured in liquid crystalline alignment medium, offer a viable alternative to traditional nuclear Overhauser enhancement (NOE)-based approaches for structure determination of large protein complexes. RDCs provide a way to constrain the relative orientation of two molecules in complex with each other by aligning their independently determined order tensors. The potential for utilization of RDCs can be extended to proteins with multiple ligands or even multimeric protein-ligand complexes, where symmetry properties of the protein can be taken advantage of. Availability of effective RDC data collection and analysis protocols can certainly aid this process by their incorporation into structure calculation protocols using intramolecular and intermolecular orientational restraints. This chapter discusses in detail some of these protocols including methods for sample preparation in liquid crystalline media, NMR experiments for RDC data collection, as well as software tools for RDC data analysis and protein-ligand complex structure determination.

Protein–protein interaction and quaternary structure

Protein–protein recognition plays an essential role in structure and function. Specific non-covalent interactions stabilize the structure of macromolecular assemblies, exemplified in this review by oligomeric proteins and the capsids of icosahedral viruses. They also allow proteins to form complexes that have a very wide range of stability and lifetimes and are involved in all cellular processes. We present some of the structure-based computational methods that have been developed to characterize the quaternary structure of oligomeric proteins and other molecular assemblies and analyze the properties of the interfaces between the subunits. We compare the size, the chemical and amino acid compositions and the atomic packing of the subunit interfaces of protein–protein complexes, oligomeric proteins, viral capsids and protein–nucleic acid complexes. These biologically significant interfaces are generally close-packed, whereas the non-specific interfaces between molecules in protein crystals are loosely packed, an observation that gives a structural basis to specific recognition. A distinction is made within each interface between a core that contains buried atoms and a solvent accessible rim. The core and the rim differ in their amino acid composition and their conservation in evolution, and the distinction helps correlating the structural data with the results of site-directed mutagenesis and in vitro studies of self-assembly.

Molecular basis of the interaction between complement receptor type 2 (CR2/CD21) and Epstein-Barr virus glycoprotein gp350

The binding of the Epstein-Barr virus glycoprotein gp350 by complement receptor type 2 (CR2) is critical for viral attachment to B lymphocytes. We set out to test hypotheses regarding the molecular nature of this interaction by developing an enzyme-linked immunosorbent assay (ELISA) for the efficient analysis of the gp350-CR2 interaction by utilizing wild-type and mutant forms of recombinant gp350 and also of the CR2 N-terminal domains SCR1 and SCR2 (designated CR2 SCR1-2). To delineate the CR2-binding site on gp350, we generated 17 gp350 single-site substitutions targeting an area of gp350 that has been broadly implicated in the binding of both CR2 and the major inhibitory anti-gp350 monoclonal antibody (MAb) 72A1. These site-directed mutations identified a novel negatively charged CR2-binding surface described by residues Glu-21, Asp-22, Glu-155, Asp-208, Glu-210, and Asp-296. We also identified gp350 amino acid residues involved in non-charge-dependent interactions with CR2, including Tyr-151, Ile-160, and Trp-162. These data were supported by experiments in which phycoerythrin-conjugated wild-type and mutant forms of gp350 were incubated with CR2-expressing K562 cells and binding was assessed by flow cytometry. The ELISA was further utilized to identify several positively charged residues (Arg-13, Arg-28, Arg-36, Lys-41, Lys-57, Lys-67, Arg-83, and Arg-89) within SCR1-2 of CR2 that are involved in the binding interaction with gp350. These experiments allowed a comparison of those CR2 residues that are important for binding gp350 to those that define the epitope for an effective inhibitory anti-CR2 MAb, 171 (Asn-11, Arg-13, Ser-32, Thr-34, Arg-36, and Tyr-64). The mutagenesis data were used to calculate a model of the CR2-gp350 complex using the soft-docking program HADDOCK.

Conformer selection and induced fit in flexible backbone protein-protein docking using computational and NMR ensembles

Accommodating backbone flexibility continues to be the most difficult challenge in computational docking of protein-protein complexes. Towards that end, we simulate four distinct biophysical models of protein binding in RosettaDock, a multiscale Monte-Carlo-based algorithm that uses a quasi-kinetic search process to emulate the diffusional encounter of two proteins and to identify low-energy complexes. The four binding models are as follows: (1) key-lock (KL) model, using rigid-backbone docking; (2) conformer selection (CS) model, using a novel ensemble docking algorithm; (3) induced fit (IF) model, using energy-gradient-based backbone minimization; and (4) combined conformer selection/induced fit (CS/IF) model. Backbone flexibility was limited to the smaller partner of the complex, structural ensembles were generated using Rosetta refinement methods, and docking consisted of local perturbations around the complexed conformation using unbound component crystal structures for a set of 21 target complexes. The lowest-energy structure contained >30% of the native residue-residue contacts for 9, 13, 13, and 14 targets for KL, CS, IF, and CS/IF docking, respectively. When applied to 15 targets using nuclear magnetic resonance ensembles of the smaller protein, the lowest-energy structure recovered at least 30% native residue contacts in 3, 8, 4, and 8 targets for KL, CS, IF, and CS/IF docking, respectively. CS/IF docking of the nuclear magnetic resonance ensemble performed equally well or better than KL docking with the unbound crystal structure in 10 of 15 cases. The marked success of CS and CS/IF docking shows that ensemble docking can be a versatile and effective method for accommodating conformational plasticity in docking and serves as a demonstration for the CS theory--that binding-competent conformers exist in the unbound ensemble and can be selected based on their favorable binding energies.

Structural deformation upon protein-protein interaction: a structural alphabet approach

Background

In a number of protein-protein complexes, the 3D structures of bound and unbound partners significantly differ, supporting the induced fit hypothesis for protein-protein binding.

Results

In this study, we explore the induced fit modifications on a set of 124 proteins available in both bound and unbound forms, in terms of local structure. The local structure is described thanks to a structural alphabet of 27 structural letters that allows a detailed description of the backbone. Using a control set to distinguish induced fit from experimental error and natural protein flexibility, we show that the fraction of structural letters modified upon binding is significantly greater than in the control set (36% versus 28%). This proportion is even greater in the interface regions (41%). Interface regions preferentially involve coils. Our analysis further reveals that some structural letters in coil are not favored in the interface. We show that certain structural letters in coil are particularly subject to modifications at the interface, and that the severity of structural change also varies. These information are used to derive a structural letter substitution matrix that summarizes the local structural changes observed in our data set. We also illustrate the usefulness of our approach to identify common binding motifs in unrelated proteins.

Conclusion

Our study provides qualitative information about induced fit. These results could be of help for flexible docking.

Incorporating biochemical information and backbone flexibility in RosettaDock for CAPRI rounds 6-12

In CAPRI rounds 6-12, RosettaDock successfully predicted 2 of 5 unbound-unbound targets to medium accuracy. Improvement over the previous method was achieved with computational mutagenesis to select decoys that match the energetics of experimentally determined hot spots. In the case of Target 21, Orc1/Sir1, this resulted in a successful docking prediction where RosettaDock alone or with simple site constraints failed. Experimental information also helped limit the interacting region of TolB/Pal, producing a successful prediction of Target 26. In addition, we docked multiple loop conformations for Target 20, and we developed a novel flexible docking algorithm to simultaneously optimize backbone conformation and rigid-body orientation to generate a wide diversity of conformations for Target 24. Continued challenges included docking of homology targets that differ substantially from their template (sequence identity <50%) and accounting for large conformational changes upon binding. Despite a larger number of unbound-unbound and homology model binding targets, Rounds 6-12 reinforced that RosettaDock is a powerful algorithm for predicting bound complex structures, especially when combined with experimental data.

A holistic approach to protein docking

Docking of unbound protein structures into a complex has gained significant progress in recent years, but nonetheless still poses a great challenge. We have pursued a holistic approach to docking which brings together effective methods at different stages. First, protein-protein interaction sites are predicted or obtained from experimental studies in the literature. Interface prediction/experimental data are then used to guide the generation of docked poses or to rank docked poses generated from an unbiased search. Finally, selected models are refined by lengthy molecular dynamics (MD) simulations in explicit water. For CAPRI target T27, we used information on interaction sites as input to drive docking and as a filter to rank docked poses. Lead candidates were then clustered according to RMSD among them. From the clustering, 10 models were selected and subject to refinement by MD simulations. Our Model 7 is rated number one among all submissions according to L_rmsd. Six of our other submissions are rated acceptable. As scorer, eight of our submissions are rated acceptable.

HADDOCK versus HADDOCK: new features and performance of HADDOCK2.0 on the CAPRI targets

Here we present version 2.0 of HADDOCK, which incorporates considerable improvements and new features. HADDOCK is now able to model not only protein-protein complexes but also other kinds of biomolecular complexes and multi-component (N > 2) systems. In the absence of any experimental and/or predicted information to drive the docking, HADDOCK now offers two additional ab initio docking modes based on either random patch definition or center-of-mass restraints. The docking protocol has been considerably improved, supporting among other solvated docking, automatic definition of semi-flexible regions, and inclusion of a desolvation energy term in the scoring scheme. The performance of HADDOCK2.0 is evaluated on the targets of rounds 4-11, run in a semi-automated mode using the original information we used in our CAPRI submissions. This enables a direct assessment of the progress made since the previous versions. Although HADDOCK performed very well in CAPRI (65% and 71% success rates, overall and for unbound targets only, respectively), a substantial improvement was achieved with HADDOCK2.0.

Protein-protein docking: progress in CAPRI rounds 6-12 using a combination of methods: the introduction of steered solvated molecular dynamics

In recent rounds of CAPRI, the Bii group has employed a combination of techniques for the prediction of the structure of protein-protein complexes. We currently use third-party software for rigid-body and semiflexible docking (MolFit, 3D-Dock, RosettaDock), and our own steered molecular dynamics (SMD) technique for flexible refinement. SMD has also been found to be useful for discriminating near-native from false positive docking decoys. In addition to this, a variety of sources of information, including multiple descriptors of interface quality combined with a QSAR-like technique, published biological information, and continuum electrostatics calculations, are also used in the assessment of candidate complexes. We shall concentrate on results for CAPRI rounds 9-11 (targets 24-27). In these rounds, the Bii group has been successful in submitting a medium quality model for each of CAPRI targets 25 and 26, and a model of acceptable quality for target 27.

M-TASSER: an algorithm for protein quaternary structure prediction

In a cell, it has been estimated that each protein on average interacts with roughly 10 others, resulting in tens of thousands of proteins known or suspected to have interaction partners; of these, only a tiny fraction have solved protein structures. To partially address this problem, we have developed M-TASSER, a hierarchical method to predict protein quaternary structure from sequence that involves template identification by multimeric threading, followed by multimer model assembly and refinement. The final models are selected by structure clustering. M-TASSER has been tested on a benchmark set comprising 241 dimers having templates with weak sequence similarity and 246 without multimeric templates in the dimer library. Of the total of 207 targets predicted to interact as dimers, 165 (80%) were correctly assigned as interacting with a true positive rate of 68% and a false positive rate of 17%. The initial best template structures have an average root mean-square deviation to native of 5.3, 6.7, and 7.4 A for the monomer, interface, and dimer structures. The final model shows on average a root mean-square deviation improvement of 1.3, 1.3, and 1.5 A over the initial template structure for the monomer, interface, and dimer structures, with refinement evident for 87% of the cases. Thus, we have developed a promising approach to predict full-length quaternary structure for proteins that have weak sequence similarity to proteins of solved quaternary structure.

meta-PPISP: a meta web server for protein-protein interaction site prediction

A number of complementary methods have been developed for predicting protein-protein interaction sites. We sought to increase prediction robustness and accuracy by combining results from different predictors, and report here a meta web server, meta-PPISP, that is built on three individual web servers: cons-PPISP (http://pipe.scs.fsu.edu/ppisp.html), Promate (http://bioportal.weizmann.ac.il/promate), and PINUP (http://sparks.informatics.iupui.edu/PINUP/). A linear regression method, using the raw scores of the three servers as input, was trained on a set of 35 nonhomologous proteins. Cross validation showed that meta-PPISP outperforms all the three individual servers. At coverages identical to those of the individual methods, the accuracy of meta-PPISP is higher by 4.8 to 18.2 percentage points. Similar improvements in accuracy are also seen on CAPRI and other targets.

AVAILABILITY

meta-PPISP can be accessed at http://pipe.scs.fsu.edu/meta-ppisp.html