Computational prediction of protein-protein complexes

ABSTRACT: BACKGROUND: Protein-protein interactions form the core of several biological processes. With protein-protein interfaces being considered as drug targets, studies on their interactions and molecular mechanisms are gaining ground. As the number of protein complexes in databases is scarce as compared to a spectrum of independent protein molecules, computational approaches are being considered for speedier model derivation and assessment of a plausible complex. In this study, a good approach towards in silico generation of protein-protein heterocomplex and identification of the most probable complex among thousands of complexes thus generated is documented. This approach becomes even more useful in the event of little or no binding site information between the interacting protein molecules. FINDINGS: A plausible protein-protein hetero-complex was fished out from 10 docked complexes which are a representative set of complexes obtained after clustering of 2000 generated complexes using protein-protein docking softwares. The interfacial area for this complex was predicted by two hotspot prediction programs employing different algorithms. Further, this complex had the lowest energy and most buried surface area of all the complexes with the same interfacial residues. CONCLUSIONS: For the generation of a plausible protein heterocomplex, various software tools were employed. Prominent are the protein-protein docking methods, prediction of 'hotspots' which are the amino acid residues likely to be in an interface and measurement of buried surface area of the complexes. Consensus generated in their predictions lends credence to the use of the various softwares used.     

An orientation-dependent hydrogen bonding potential improves prediction of specificity and structure for proteins and protein-protein complexes

Hydrogen bonding is a key contributor to the specificity of intramolecular and intermolecular interactions in biological systems. Here, we develop an orientation-dependent hydrogen bonding potential based on the geometric characteristics of hydrogen bonds in high-resolution protein crystal structures, and evaluate it using four tests related to the prediction and design of protein structures and protein-protein complexes. The new potential is superior to the widely used Coulomb model of hydrogen bonding in prediction of the sequences of proteins and protein-protein interfaces from their structures, and improves discrimination of correctly docked protein-protein complexes from large sets of alternative structures.     

Equilibrium amide hydrogen exchange and protein folding kinetics

The classical Linderstrøm-Lang hydrogen exchange (HX) model is extended to describe the relationship between the HX behaviors (EX1 and EX2) and protein folding kinetics for the amide protons that can only exchange by global unfolding in a three-state system including native (N), intermediate (I), and unfolded (U) states. For these slowly exchanging amide protons, it is shown that the existence of an intermediate (I) has no effect on the HX behavior in an off-pathway three-state system (IUN). On the other hand, in an on-pathway three-state system (UIN), the existence of a stable folding intermediate has profound effect on the HX behavior. It is shown that fast refolding from the unfolded state to the stable intermediate state alone does not guarantee EX2 behavior. The rate of refolding from the intermediate state to the native state also plays a crucial role in determining whether EX1 or EX2 behavior should occur. This is mainly due to the fact that only amide protons in the native state are observed in the hydrogen exchange experiment. These new concepts suggest that caution needs to be taken if one tries to derive the kinetic events of protein folding from equilibrium hydrogen exchange experiments.     

Kinetics of unfolding and folding from amide hydrogen exchange in native ubiquitin

Amide hydrogen (NH) exchange is one of the few experimental techniques with the potential for determining the thermodynamics and kinetics of conformational motions at nearly every residue in native proteins. Quantitative interpretation of NH exchange in terms of molecular motions relies on a simple two-state kinetic model: at any given slowly exchanging NH, a closed or exchange-incompetent conformation is in equilibrium with an open or exchange-competent conformation. Previous studies have demonstrated the accuracy of this model in measuring conformational equilibria by comparing exchange data with the thermodynamics of protein unfolding. We report here a test of the accuracy of the model in determining the kinetics of conformational changes in native proteins. The kinetics of folding and unfolding for ubiquitin have been measured by conventional methods and compared with those derived from a comprehensive analysis of the pH dependence of exchange in native ubiquitin. Rate constants for folding and unfolding from these two very different types of experiments show good agreement. The simple model for NH exchange thus appears to be a robust framework for obtaining quantitative information about molecular motions in native proteins.     

Partner-aware prediction of interacting residues in protein-protein complexes from sequence data

Computational prediction of residues that participate in protein-protein interactions is a difficult task, and state of the art methods have shown only limited success in this arena. One possible problem with these methods is that they try to predict interacting residues without incorporating information about the partner protein, although it is unclear how much partner information could enhance prediction performance. To address this issue, the two following comparisons are of crucial significance: (a) comparison between the predictability of inter-protein residue pairs, i.e., predicting exactly which residue pairs interact with each other given two protein sequences; this can be achieved by either combining conventional single-protein predictions or making predictions using a new model trained directly on the residue pairs, and the performance of these two approaches may be compared: (b) comparison between the predictability of the interacting residues in a single protein (irrespective of the partner residue or protein) from conventional methods and predictions converted from the pair-wise trained model. Using these two streams of training and validation procedures and employing similar two-stage neural networks, we showed that the models trained on pair-wise contacts outperformed the partner-unaware models in predicting both interacting pairs and interacting single-protein residues. Prediction performance decreased with the size of the conformational change upon complex formation; this trend is similar to docking, even though no structural information was used in our prediction. An example application that predicts two partner-specific interfaces of a protein was shown to be effective, highlighting the potential of the proposed approach. Finally, a preliminary attempt was made to score docking decoy poses using prediction of interacting residue pairs; this analysis produced an encouraging result.     

Comparisons of pressure and temperature activation parameters for amide hydrogen exchange in T4 lysozyme

Activation enthalpies and entropies are reported for proton-deuteron exchange at 42 amide sites in T4 lysozyme and compared with activation volumes for the same residues obtained earlier [Hitchens, T. K., and Bryant, R. G. (1998) Biochemistry 37, 5878-5887]. There is no correlation found between activation volume and activation entropy or activation enthalpy. The activation enthalpy is linearly related to the activation entropy in part as a consequence of a relatively narrow sampling window for the rate constants that corresponds to a narrow range of activation free energy. A consequence of the entropy-enthalpy compensation is preservation of rank order of proton exchange. Variations in DeltaH, DeltaS, and DeltaV for residues that are structurally close together in the folded protein suggest that there may be a variety of energetically distinct pathways for the access of solvent to these structurally related exchange sites.     

Backbone dynamics of detergent-solubilized alamethicin from amide hydrogen exchange measurements.

Alamethicin is a 20 amino acid antibiotic peptide produced by the soil fungus Trichoderma viride. The peptide inserts into bacterial membranes and self-associates to form ion channels, but the details of this process are unknown. Residue-specific acid- and base-catalyzed exchange data were obtained for 16 of 18 backbone amides of alamethicin dissolved in sodium dodecyl sulfate micelles using high-resolution 2-dimensional heteronuclear nuclear magnetic resonance spectroscopy. To facilitate interpretation of the exchange data, we synthesized N-acetyl-alpha-aminoisobutyric acid-N'-methyl and N-acetyl-alanine-N'-methyl and measured the pD dependence of their hydrogen-deuterium exchange rates to determine the sequence-dependent inductive and steric effects of the alpha-aminoisobutyric acid residue. Intramolecular H-bonding in alamethicin was monitored through the exchange parameters kmin (minimum exchange rate) which indicate that the backbone is significantly more stable than the backbones of alanine-based helical peptides. Rapid exchange at Gly-11 suggests a highly local conformational flexibility in the middle of the peptide. Interactions with the detergent micelle were revealed by the exchange parameters pDmin (pD of minimum exchange) which suggest that the N-terminus of alamethicin interacts more strongly with the detergent micelle than does the C-terminus. A periodicity in pDmin difference data reveals that one surface of the helix interacts more strongly with the micelle. The surface consists of residues 1, 5, 9, 13, 16, and 20. The opposite face of the helix contains several polar residues (two glutamines and a glycine), suggesting that, on average, this face of the helix is directed toward the solvent. These results serve as a model for the interaction of the peptide with membranes containing anionic lipid. In combination with published molecular dynamics simulations [Gibbs et al. (1997) Biophys. J. 72, 2490-2495], the present results also offer insight into the mechanisms of hydrogen-deuterium exchange in helical peptides.     

Effects of co-operative ligand binding on protein amide NH hydrogen exchange

Amide protection factors have been determined from NMR measurements of hydrogen/deuterium amide NH exchange rates measured on assigned signals from Lactobacillus casei apo-DHFR and its binary and ternary complexes with trimethoprim (TMP), folinic acid and coenzymes (NADPH/NADP(+)). The substantial sizes of the residue-specific DeltaH and TDeltaS values for the opening/closing events in NH exchange for most of the measurable residues in apo-DHFR indicate that sub-global or global rather than local exchange mechanisms are usually involved. The amide groups of residues in helices and sheets are those most protected in apo-DHFR and its complexes, and the protection factors are generally related to the tightness of ligand binding. The effects of ligand binding that lead to changes in amide protection are not localised to specific binding sites but are spread throughout the structure via a network of intramolecular interactions. Although the increase in protein stability in the DHFR.TMP.NADPH complex involves increased ordering in the protein structure (requiring TDeltaS energy) this is recovered, to a large extent, by the stronger binding (enthalpic DeltaH) interactions made possible by the reduced motion in the protein. The ligand-induced protection effects in the ternary complexes DHFR.TMP.NADPH (large positive binding co-operativity) and DHFR.folinic acid.NADPH (large negative binding co-operativity) mirror the co-operative effects seen in the ligand binding. For the DHFR.TMP.NADPH complex, the ligand-induced protection factors result in DeltaDeltaG(o) values for many residues being larger than the DeltaDeltaG(o) values in the corresponding binary complexes. In contrast, for DHFR.folinic acid.NADPH, the DeltaDeltaG(o) values are generally smaller than many of those in the corresponding binary complexes. The results indicate that changes in protein conformational flexibility on formation of the ligand complex play an important role in determining the co-operativity in the ligand binding.     

Helix bending in alamethicin: molecular dynamics simulations and amide hydrogen exchange in methanol

Molecular dynamics simulations of alamethicin in methanol were carried out with either a regular alpha-helical conformation or the x-ray crystal structure as starting structures. The structures rapidly converged to a well-defined hydrogen-bonding pattern with mixed alpha-helical and 3(10)-helical hydrogen bonds, consistent with NMR structural characterization, and did not unfold throughout the 1-ns simulation, despite some sizable backbone fluctuations involving reversible breaking of helical hydrogen bonds. Bending of the helical structure around residues Aib10-Aib13 was associated with reversible flips of the peptide bonds involving G11 (Aib10-G11 or G11-L12 peptide bonds), yielding discrete structural states in which the Aib10 carbonyl or (rarely) the G11 carbonyl was oriented away from the peptide helix. These peptide bond reversals could be accommodated without greatly perturbing the adjacent helical structure, and intramolecular hydrogen bonding was generally maintained in bent states through the formation of new (non-alpha or 3[10]) hydrogen bonds with good geometries: G11 NH-V9 CO (inverse gamma turn), Aib13 NH-Aib8 CO (pi-helix) and, rarely, L12 NH- Q7 NH (pi-helix). These observations may reconcile potentially conflicting NMR structural information for alamethicin in methanol, in which evidence for conformational flexibility in the peptide sequence before P14 (G11-Aib13) contrasts with the stability of backbone amide NH groups to exchange with solvent. Similar reversible reorientation of the Thr11-Gly12 peptide bond of melittin is also observed in dynamics simulations in methanol (R. B. Sessions, N. Gibbs, and C. E. Dempsey, submitted). This phenomenon may have some role in the orientation of the peptide carbonyl in solvating the channel lumen in membrane ion channel states of these peptides.     

Measurement of amide hydrogen exchange rates with the use of radiation damping

A simple method for measuring amide hydrogen exchange rates is presented, which is based on the selective inversion of water magnetization with the use of radiation damping. Simulations show that accurate exchange rates can be measured despite the complications of radiation damping and cross relaxation to the exchange process between amide and water protons. This method cannot eliminate the contributions of the exchange-relayed NOE and direct NOE to the measured exchange rates, but minimize the direct NOE contribution. In addition, the amides with a significant amount of such indirect contributions are possible to be identified from the shape of the exchange peak intensity profiles or/and from the apparent relaxation rates of amide protons which are extracted from fitting the intensity profiles to an equation established here for our experiment. The method was tested on ubiquitin and also applied to an acyl carrier protein. The amide exchange rates for the acyl carrier protein at two pHs indicate that the entire protein is highly dynamic on the second timescale. Low protection factors for the residues in the regular secondary structural elements also suggest the presence of invisible unfolded species. The highly dynamic nature of the acyl carrier protein may be crucial for its interactions with its substrate and enzymes.     

Thermal-induced unfolding domains in aldolase identified by amide hydrogen exchange and mass spectrometry.

Amide hydrogen exchange has been measured in short segments of intact rabbit muscle aldolase at temperatures of 14-50 degrees C by the protein fragmentation/mass spectrometry method (Zhang Z, Smith DL, 1993, Protein Sci 2:522-531). Deuterium levels in some segments did not change over the temperature range of the measurements, whereas deuterium levels in other segments increased rapidly with temperature. These results demonstrate that the equilibrium constant for local unfolding, Kunf, of some segments increases with temperature in the low temperature range (14-30 degrees C) of this study. Aldolase begins to lose activity at temperatures above 40 degrees C. In the 40-50 degrees C temperature range, Kunf is greater than 10(-4) in some regions and less than 10(-6) in other regions. This wide range of regional stability in the temperature range where aldolase begins to denature is interpreted in terms of cooperative unfolding/folding domains. Regions of highest stability were located along the hydrophobic subunit binding surface. It is proposed that hydrogen exchange might be used to identify unfolding domains in multidomain proteins whose thermodynamic properties have been determined by differential scanning calorimetry.     

Conformational analysis of Epac activation using amide hydrogen/deuterium exchange mass spectrometry

Exchange proteins directly activated by cAMP (Epac) play important roles in mediating the effects of cAMP through the activation of downstream small GTPases, Rap. To delineate the mechanism of Epac activation, we probed the conformation and structural dynamics of Epac using amide hydrogen/deuterium exchange and structural modeling. Our studies show that cAMP induces significant conformational changes that lead to a spatial rearrangement of the regulatory components of Epac and allows the exposure of the catalytic core for effector binding without imposing significant conformational change on the catalytic core. Homology modeling and comparative structural analyses of the cAMP binding domains of Epac and cAMP-dependent protein kinase (PKA) lead to a model of Epac activation, in which Epac and PKA activation by cAMP employs the same underlying principle, although the detailed structural and conformational changes associated with Epac and PKA activation are significantly different.     

Using indirect protein-protein interactions for protein complex prediction

Protein complexes are fundamental for understanding principles of cellular organizations. As the sizes of protein-protein interaction (PPI) networks are increasing, accurate and fast protein complex prediction from these PPI networks can serve as a guide for biological experiments to discover novel protein complexes. However, it is not easy to predict protein complexes from PPI networks, especially in situations where the PPI network is noisy and still incomplete. Here, we study the use of indirect interactions between level-2 neighbors (level-2 interactions) for protein complex prediction. We know from previous work that proteins which do not interact but share interaction partners (level-2 neighbors) often share biological functions. We have proposed a method in which all direct and indirect interactions are first weighted using topological weight (FS-Weight), which estimates the strength of functional association. Interactions with low weight are removed from the network, while level-2 interactions with high weight are introduced into the interaction network. Existing clustering algorithms can then be applied to this modified network. We have also proposed a novel algorithm that searches for cliques in the modified network, and merge cliques to form clusters using a "partial clique merging" method. Experiments show that (1) the use of indirect interactions and topological weight to augment protein-protein interactions can be used to improve the precision of clusters predicted by various existing clustering algorithms; and (2) our complex-finding algorithm performs very well on interaction networks modified in this way. Since no other information except the original PPI network is used, our approach would be very useful for protein complex prediction, especially for prediction of novel protein complexes.     

Electrostatic properties of protein-protein complexes

Statistical electrostatic analysis of 37 protein-protein complexes extracted from the previously developed database of protein complexes (ProtCom, http://www.ces.clemson.edu/compbio/protcom) is presented. It is shown that small interfaces have a higher content of charged and polar groups compared to large interfaces. In a vast majority of the cases the average pKa shifts for acidic residues induced by the complex formation are negative, indicating that complex formation stabilizes their ionizable states, whereas the histidines are predicted to destabilize the complex. The individual pKa shifts show the same tendency since 80% of the interfacial acidic groups were found to lower their pKas, whereas only 25% of histidines raise their pKa upon the complex formation. The interfacial groups have been divided into three sets according to the mechanism of their pKa shift, and statistical analysis of each set was performed. It was shown that the optimum pH values (pH of maximal stability) of the complex tend to be the same as the optimum pH values of the complex components. This finding can be used in the homology-based prediction of the 3D structures of protein complexes, especially when one needs to evaluate and rank putative models. It is more likely for a model to be correct if both components of the model complex and the entire complex have the same or at least similar values of the optimum pH.     

Choosing negative examples for the prediction of protein-protein interactions

The protein-protein interaction networks of even well-studied model organisms are sketchy at best, highlighting the continued need for computational methods to help direct experimentalists in the search for novel interactions. This need has prompted the development of a number of methods for predicting protein-protein interactions based on various sources of data and methodologies. The common method for choosing negative examples for training a predictor of protein-protein interactions is based on annotations of cellular localization, and the observation that pairs of proteins that have different localization patterns are unlikely to interact. While this method leads to high quality sets of non-interacting proteins, we find that this choice can lead to biased estimates of prediction accuracy, because the constraints placed on the distribution of the negative examples makes the task easier. The effects of this bias are demonstrated in the context of both sequence-based and non-sequence based features used for predicting protein-protein interactions.     

Mapping of discontinuous conformational epitopes by amide hydrogen/deuterium exchange mass spectrometry and computational docking

Understanding antigen-antibody interactions at the sub-molecular level is of particular interest for scientific, regulatory, and intellectual property reasons, especially with increasing demand for monoclonal antibody therapeutic agents. Although various techniques are available for the determination of an epitope, there is no widely applicable, high-resolution, and reliable method available. Here, a combination approach using amide hydrogen/deuterium exchange coupled with proteolysis and mass spectrometry (HDX-MS) and computational docking was applied to investigate antigen-antibody interactions. HDX-MS is a widely applicable, medium-resolution, medium-throughput technology that can be applied to epitope identification. First, the epitopes of cytochrome c-E8, IL-13-CNTO607, and IL-17A-CAT-2200 interactions identified using the HDX-MS method were compared with those identified by X-ray co-crystal structures. The identified epitopes are in good agreement with those identified using high-resolution X-ray crystallography. Second, the HDX-MS data were used as constraints for computational docking. More specifically, the non-epitope residues of an antigen identified using HDX-MS were designated as binding ineligible during computational docking. This approach, termed HDX-DOCK, gave more tightly clustered docking poses than stand-alone docking for all antigen-antibody interactions examined and improved docking results significantly for the cytochrome c-E8 interaction.     

Optimization of electrostatic interactions in protein-protein complexes

In this article, we present a statistical analysis of the electrostatic properties of 298 protein-protein complexes and 356 domain-domain structures extracted from the previously developed database of protein complexes (ProtCom, http://www.ces.clemson.edu/compbio/protcom). For each structure in the dataset we calculated the total electrostatic energy of the binding and its two components, Coulombic and reaction field energy. It was found that in a vast majority of the cases (>90%), the total electrostatic component of the binding energy was unfavorable. At the same time, the Coulombic component of the binding energy was found to favor the complex formation while the reaction field component of the binding energy opposed the binding. It was also demonstrated that the components in a wild-type (WT) structure are optimized/anti-optimized with respect to the corresponding distributions, arising from random shuffling of the charged side chains. The degree of this optimization was assessed through the Z-score of WT energy in respect to the random distribution. It was found that the Z-scores of Coulombic interactions peak at a considerably negative value for all 654 cases considered while the Z-score of the reaction field energy varied among different types of complexes. All these findings indicate that the Coulombic interactions within WT protein-protein complexes are optimized to favor the complex formation while the total electrostatic energy predominantly opposes the binding. This observation was used to discriminate WT structures among sets of structural decoys and showed that the electrostatic component of the binding energy is not a good discriminator of the WT; while, Coulombic or reaction field energies perform better depending upon the decoy set used.     

Computational prediction of host-pathogen protein-protein interactions

MOTIVATION: Infectious diseases such as malaria result in millions of deaths each year. An important aspect of any host-pathogen system is the mechanism by which a pathogen can infect its host. One method of infection is via protein-protein interactions (PPIs) where pathogen proteins target host proteins. Developing computational methods that identify which PPIs enable a pathogen to infect a host has great implications in identifying potential targets for therapeutics. RESULTS: We present a method that integrates known intra-species PPIs with protein-domain profiles to predict PPIs between host and pathogen proteins. Given a set of intra-species PPIs, we identify the functional domains in each of the interacting proteins. For every pair of functional domains, we use Bayesian statistics to assess the probability that two proteins with that pair of domains will interact. We apply our method to the Homo sapiens-Plasmodium falciparum host-pathogen system. Our system predicts 516 PPIs between proteins from these two organisms. We show that pairs of human proteins we predict to interact with the same Plasmodium protein are close to each other in the human PPI network and that Plasmodium pairs predicted to interact with same human protein are co-expressed in DNA microarray datasets measured during various stages of the Plasmodium life cycle. Finally, we identify functionally enriched sub-networks spanned by the predicted interactions and discuss the plausibility of our predictions. AVAILABILITY: Supplementary data are available at http://staff.vbi.vt.edu/dyermd/publications/dyer2007a.html. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.