Small-molecule inhibitor starting points learned from protein-protein interaction inhibitor structure

MOTIVATION

Protein-protein interactions (PPIs) are a promising, but challenging target for pharmaceutical intervention. One approach for addressing these difficult targets is the rational design of small-molecule inhibitors that mimic the chemical and physical properties of small clusters of key residues at the protein-protein interface. The identification of appropriate clusters of interface residues provides starting points for inhibitor design and supports an overall assessment of the susceptibility of PPIs to small-molecule inhibition.

RESULTS

We extract Small-Molecule Inhibitor Starting Points (SMISPs) from protein-ligand and protein-protein complexes in the Protein Data Bank (PDB). These SMISPs are used to train two distinct classifiers, a support vector machine and an easy to interpret exhaustive rule classifier. Both classifiers achieve better than 70% leave-one-complex-out cross-validation accuracy and correctly predict SMISPs of known PPI inhibitors not in the training set. A PDB-wide analysis suggests that nearly half of all PPIs may be susceptible to small-molecule inhibition.

Community-wide assessment of protein-interface modeling suggests improvements to design methodology

The CAPRI (Critical Assessment of Predicted Interactions) and CASP (Critical Assessment of protein Structure Prediction) experiments have demonstrated the power of community-wide tests of methodology in assessing the current state of the art and spurring progress in the very challenging areas of protein docking and structure prediction. We sought to bring the power of community-wide experiments to bear on a very challenging protein design problem that provides a complementary but equally fundamental test of current understanding of protein-binding thermodynamics. We have generated a number of designed protein-protein interfaces with very favorable computed binding energies but which do not appear to be formed in experiments, suggesting that there may be important physical chemistry missing in the energy calculations. A total of 28 research groups took up the challenge of determining what is missing: we provided structures of 87 designed complexes and 120 naturally occurring complexes and asked participants to identify energetic contributions and/or structural features that distinguish between the two sets. The community found that electrostatics and solvation terms partially distinguish the designs from the natural complexes, largely due to the nonpolar character of the designed interactions. Beyond this polarity difference, the community found that the designed binding surfaces were, on average, structurally less embedded in the designed monomers, suggesting that backbone conformational rigidity at the designed surface is important for realization of the designed function. These results can be used to improve computational design strategies, but there is still much to be learned; for example, one designed complex, which does form in experiments, was classified by all metrics as a nonbinder.

Pharmer: efficient and exact pharmacophore search

Pharmacophore search is a key component of many drug discovery efforts. Pharmer is a new computational approach to pharmacophore search that scales with the breadth and complexity of the query, not the size of the compound library being screened. Two novel methods for organizing pharmacophore data, the Pharmer KDB-tree and Bloom fingerprints, enable Pharmer to perform an exact pharmacophore search of almost two million structures in less than a minute. In general, Pharmer is more than an order of magnitude faster than existing technologies. The complete source code is available under an open-source license at http://pharmer.sourceforge.net .

Novel modulation factor quantifies the role of water molecules in protein interactions

Water molecules decrease the potential of mean force of a hydrogen bond (H-bond), as well as modulate (de)solvation forces, but exactly how much has not been easy to determine. Crystallographic water molecules provide snapshots of optimal solutions for the role of solvent in protein interactions, information that is often ignored by implicit solvent models. Motivated by high-resolution crystal structures, we describe a simple quantitative approach to explicitly incorporate the role of molecular water in protein interactions. Applications to protein-DNA interactions show that the accuracy of binding free-energy estimates improves significantly if a distinction is made between H-bonds that are desolvated (or only contact crystal waters), solvated by mobile waters trapped at the binding interface, or partially solvated through connections to bulk water. These different environments are modeled by a unique "water" scaling factor that decreases or increases the strength of hydrogen bonds depending on whether water contacts the acceptor or donor atoms or the bond is fully desolvated, respectively. Our empirical energies are fully consistent with mobile water molecules having a strong polarization effect in direct intermolecular interactions.

ANCHOR: a web server and database for analysis of protein-protein interaction binding pockets for drug discovery

ANCHOR is a web-based tool whose aim is to facilitate the analysis of protein–protein interfaces with regard to its suitability for small molecule drug design. To this end, ANCHOR exploits the so-called anchor residues, i.e. amino acid side-chains deeply buried at protein–protein interfaces, to indicate possible druggable pockets to be targeted by small molecules. For a given protein–protein complex submitted by the user, ANCHOR calculates the change in solvent accessible surface area (ΔSASA) upon binding for each side-chain, along with an estimate of its contribution to the binding free energy. A Jmol-based tool allows the user to interactively visualize selected anchor residues in their pockets as well as the stereochemical properties of the surrounding region such as hydrogen bonding. ANCHOR includes a Protein Data Bank (PDB) wide database of pre-computed anchor residues from more than 30 000 PDB entries with at least two protein chains. The user can query according to amino acids, buried area (SASA), energy or keywords related to indication areas, e.g. oncogene or diabetes. This database provides a resource to rapidly assess protein–protein interactions for the suitability of small molecules or fragments with bioisostere anchor analogues as possible compounds for pharmaceutical intervention. ANCHOR web server and database are freely available at http://structure.pitt.edu/anchor.

Electrostatic hot spot on DNA-binding domains mediates phosphate desolvation and the pre-organization of specificity determinant side chains

A major obstacle towards elucidating the molecular basis of transcriptional regulation is the lack of a detailed understanding of the interplay between non-specific and specific protein–DNA interactions. Based on molecular dynamics simulations of C₂H₂ zinc fingers (ZFs) and engrailed homeodomain transcription factors (TFs), we show that each of the studied DNA-binding domains has a set of highly constrained side chains in preset configurations ready to form hydrogen bonds with the DNA backbone. Interestingly, those domains that bury their recognition helix into the major groove are found to have an electrostatic hot spot for Cl⁻ ions located on the same binding cavity as the most buried DNA phosphate. The spot is characterized by three protein hydrogen bond donors, often including two basic side chains. If bound, Cl⁻ ions, likely mimicking phosphates, steer side chains that end up forming specific contacts with bases into bound-like conformations. These findings are consistent with a multi-step DNA-binding mechanism in which a pre-organized set of TF side chains assist in the desolvation of phosphates into well defined sites, prompting the re-organization of specificity determining side chains into conformations suitable for the recognition of their cognate sequence.

Optimization of minimum set of protein-DNA interactions: a quasi exact solution with minimum over-fitting

Motivation: A major limitation in modeling protein interactions is the difficulty of assessing the over-fitting of the training set. Recently, an experimentally based approach that integrates crystallographic information of C2H2 zinc finger–DNA complexes with binding data from 11 mutants, 7 from EGR finger I, was used to define an improved interaction code (no optimization). Here, we present a novel mixed integer programming (MIP)-based method that transforms this type of data into an optimized code, demonstrating both the advantages of the mathematical formulation to minimize over- and under-fitting and the robustness of the underlying physical parameters mapped by the code.

Results: Based on the structural models of feasible interaction networks for 35 mutants of EGR–DNA complexes, the MIP method minimizes the cumulative binding energy over all complexes for a general set of fundamental protein–DNA interactions. To guard against over-fitting, we use the scalability of the method to probe against the elimination of related interactions. From an initial set of 12 parameters (six hydrogen bonds, five desolvation penalties and a water factor), we proceed to eliminate five of them with only a marginal reduction of the correlation coefficient to 0.9983. Further reduction of parameters negatively impacts the performance of the code (under-fitting). Besides accurately predicting the change in binding affinity of validation sets, the code identifies possible context-dependent effects in the definition of the interaction networks. Yet, the approach of constraining predictions to within a pre-selected set of interactions limits the impact of these potential errors to related low-affinity complexes.

Contact:ccamacho@pitt.edu; droleg@pitt.edu

Supplementary information:Supplementary data are available at Bioinformatics online.

SUMOylation of nuclear actin

Actin, a major component of the cytoplasm, is also abundant in the nucleus. Nuclear actin is involved in a variety of nuclear processes including transcription, chromatin remodeling, and intranuclear transport. Nevertheless, the regulation of nuclear actin by posttranslational modifications has not been investigated. We now show that nuclear actin is modified by SUMO2 and SUMO3 and that computational modeling and site-directed mutagenesis identified K68 and K284 as critical sites for SUMOylating actin. We also present a model for the actin–SUMO complex and show that SUMOylation is required for the nuclear localization of actin.

Experimentally based contact energies decode interactions responsible for protein-DNA affinity and the role of molecular waters at the binding interface

A major obstacle towards understanding the molecular basis of transcriptional regulation is the lack of a recognition code for protein–DNA interactions. Using high-quality crystal structures and binding data on the promiscuous family of C₂H₂ zinc fingers (ZF), we decode 10 fundamental specific interactions responsible for protein–DNA recognition. The interactions include five hydrogen bond types, three atomic desolvation penalties, a favorable non-polar energy, and a novel water accessibility factor. We apply this code to three large datasets containing a total of 89 C₂H₂ transcription factor (TF) mutants on the three ZFs of EGR. Guided by molecular dynamics simulations of individual ZFs, we map the interactions into homology models that embody all feasible intra- and intermolecular bonds, selecting for each sequence the structure with the lowest free energy. These interactions reproduce the change in affinity of 35 mutants of finger I (R² = 0.998), 23 mutants of finger II (R² = 0.96) and 31 finger III human domains (R² = 0.94). Our findings reveal recognition rules that depend on DNA sequence/structure, molecular water at the interface and induced fit of the C₂H₂ TFs. Collectively, our method provides the first robust framework to decode the molecular basis of TFs binding to DNA.

Structural and thermodynamic approach to peptide immunogenicity

In the conventional paradigm of humoral immunity, B cells recognize their cognate antigen target in its native form. However, it is well known that relatively unstable peptides bearing only partial structural resemblance to the native protein can trigger antibodies recognizing higher-order structures found in the native protein. On the basis of sound thermodynamic principles, this work reveals that stability of immunogenic proteinlike motifs is a critical parameter rationalizing the diverse humoral immune responses induced by different linear peptide epitopes. In this paradigm, peptides with a minimal amount of stability (ΔG_X<0 kcal/mol) around a proteinlike motif (X) are capable of inducing antibodies with similar affinity for both peptide and native protein, more weakly stable peptides (ΔG_X>0 kcal/mol) trigger antibodies recognizing full protein but not peptide, and unstable peptides (ΔG_X>8 kcal/mol) fail to generate antibodies against either peptide or protein. Immunization experiments involving peptides derived from the autoantigen histidyl-tRNA synthetase verify that selected peptides with varying relative stabilities predicted by molecular dynamics simulations induce antibody responses consistent with this theory. Collectively, these studies provide insight pertinent to the structural basis of immunogenicity and, at the same time, validate this form of thermodynamic and molecular modeling as an approach to probe the development/evolution of humoral immune responses.

In the current paradigm of immune system recognition, T cell receptors bind to relatively short peptide sequences complexed with major histocompatibility complex proteins on the surface of antigen presenting cells, while B cell receptors recognize unprocessed protein structures. Yet, ample data exist showing that peptide immunization can trigger B cell responses targeting both the immunizing peptide and peptidelike motifs contained within intact protein—despite the fact that the folding stability of such peptides is often quite low. Using thermodynamic modeling and the technique of molecular dynamics simulations, this work provides a cogent framework for understanding the relative capacity of inherently unstable peptide structures to faithfully trigger B cell antibody production against specific conformational motifs found in native/intact proteins.

Acidic groups docked to well defined wetted pockets at the core of the binding interface: a tale of scoring and missing protein interactions in CAPRI

Some challenging targets in CAPRI (T24/25 and T26) involve binding solvent accessible acidic residues at the core of the binding interface, where they are always found immersed in crystal waters. In fact, Asp and Glu residues are more likely to form part of the hydrogen bond network of their surrounding crystal water molecules than to form a buried salt bridge. Interestingly, many of the crystal waters mediating the intermolecular interactions of the acidic groups are already present in the unbound structure, reinforcing the notion that some water molecules behave as an extension of the protein structure. This is in contrast to acidic groups found in the periphery of the binding interface that form ubiquitous salt bridges that cement the high affinity complex, while at the same time they are exposed to rapidly exchanging water molecules. Because of this, dichotomy implicit solvent scoring functions fail to properly rank these complexes by prioritizing salt bridges rather than water mediated contacts. A detailed analysis of Target 24, for which our group predicted two out of the four successful homology model complex structures, and Target 26 reveal how crystal waters shape the binding cavities of acidic groups prior to binding, in agreement with the theory of anchor residues as mediators of protein recognition.

Species-specific immune responses generated by histidyl-tRNA synthetase immunization are associated with muscle and lung inflammation

Evidence implicating histidyl-tRNA synthetase (Jo-1) in the pathogenesis of the anti-synthetase syndrome includes established genetic associations linking the reproducible phenotype of muscle inflammation and interstitial lung disease with autoantibodies recognizing Jo-1. To better address the role of Jo-1-directed B and T cell responses in the context of different genetic backgrounds, we employed Jo-1 protein immunization of C57BL/6 and NOD congenic mice. Detailed analysis of early antibody responses following inoculation with human or murine Jo-1 demonstrates remarkable species-specifity, with limited cross recognition of Jo-1 from the opposite species. Complementing these results, immunization with purified peptides derived from murine Jo-1 generates B and T cells targeting species-specific epitopes contained within the amino terminal 120 amino acids of murine Jo-1. The eventual spreading of B cell epitopes that uniformly occurs 8 weeks post immunization with murine Jo-1 provides additional evidence of an immune response mediated by autoreactive, Jo-1-specific T cells. Corresponding to this self-reactivity, mice immunized with murine Jo-1 develop a striking combination of muscle and lung inflammation that replicates features of the human anti-synthetase syndrome.

Tenascin cytotactin epidermal growth factor-like repeat binds epidermal growth factor receptor with low affinity

Select epidermal growth factor (EGF)-like (EGFL) repeats of human tenascin cytotactin (tenascin C) can stimulate EGF receptor (EGFR) signaling, but activation requires micromolar concentrations of soluble EGFL repeats in contrast to subnanomolar concentrations of classical growth factors such as EGF. Using in silico homology modeling techniques, we generated a structure for one such repeat, the 14th EGFL repeat (Ten14). Ten14 assumes a tight EGF-like fold with truncated loops, consistent with circular dichroism studies. We generated bound structures for Ten14 with EGFR using two different approaches, resulting in two distinctly different conformations. Normal mode analysis of both structures indicated that the binding pocket of EGFR exhibits a significantly higher mobility in Ten14-EGFR complex compared to that of the EGF-EGFR complex; we hypothesized this may be attributed to loss of key high-affinity interactions within the Ten14-EGFR complex. We proved the efficacy of our in silico models by in vitro experiments. Surface plasmon resonance measurements yielded equilibrium constant K(D) of 74 microM for Ten14, approximately three orders of magnitude weaker than that of EGF. In accordance with our predicted bound models, Ten14 in monomeric form does not bind EGFR with sufficient stability so as to induce degradation of receptor, or undergo EGFR-mediated internalization over either the short (20 min) or long (48 h) term. This transient interaction with the receptor on the cell surface is in marked contrast to other EGFR ligands which cause EGFR transit through, and signaling from intracellular locales in addition to cell surface signaling.

FastContact: a free energy scoring tool for protein-protein complex structures

‘FastContact’ is a server that estimates the direct electrostatic and desolvation interaction free energy between two proteins in units of kcal/mol. Users submit two proteins in PDB format, and the output is emailed back to the user in three files: one output file, and the two processed proteins. Besides the electrostatic and desolvation free energy, the server reports residue contact free energies that rapidly highlight the hotspots of the interaction and evaluates the van der Waals interaction using CHARMm. Response time is ∼1 min. The server has been successfully tested and validated, scoring refined complex structures and blind sets of docking decoys, as well as proven useful predicting protein interactions. ‘FastContact’ offers unique capabilities from biophysical insights to scoring and identifying important contacts.

SIMPLE estimate of the free energy change due to aliphatic mutations: Superior predictions based on first principles

The bioinformatics revolution of the last decade has been instrumental in the development of empirical potentials to quantitatively estimate protein interactions for modeling and design. Although computationally efficient, these potentials hide most of the relevant thermodynamics in 5-to-40 parameters that are fitted against a large experimental database. Here, we revisit this longstanding problem and show that a careful consideration of the change in hydrophobicity, electrostatics, and configurational entropy between the folded and unfolded state of aliphatic point mutations predicts 20-30% less false positives and yields more accurate predictions than any published empirical energy function. This significant improvement is achieved with essentially no free parameters, validating past theoretical and experimental efforts to understand the thermodynamics of protein folding. Our first principle analysis strongly suggests that both the solute-solute van der Waals interactions in the folded state and the electrostatics free energy change of exposed aliphatic mutations are almost completely compensated by similar interactions operating in the unfolded ensemble. Not surprisingly, the problem of properly accounting for the solvent contribution to the free energy of polar and charged group mutations, as well as of mutations that disrupt the protein backbone remains open.

Ensemble modeling of substrate binding to cytochromes P450: analysis of catalytic differences between CYP1A orthologs

A novel application of modeling and docking approaches involving ensembles of homology models is used to understand structural bases underlying subtle catalytic differences between related cytochromes P450 (CYPs). Mammalian CYP1A1s and fish CYP1As are orthologous enzymes with similar substrate preferences. With some substrates (3,3',4,4'-tetrachlorobiphenyl, TCB) oxidation rates differ by orders of magnitude, while others (e.g., benzo[a]pyrene; B[a]P) are oxidized at similar rates but with somewhat differing regiospecificity. These two environmental chemical substrates (TCB and B[a]P) as well as 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) were docked to multiple models of rat, human, scup, and/or killifish CYP1As, based on multiple templates, retaining multiple poses from each model, giving ensembles of docked poses for each species. With TCB, more poses were observed closer to the heme in ensembles of rat or human CYP1A1 than of killifish CYP1A. Analysis of interacting residues suggested that differences in TCB pose distributions are due primarily to Leu387 and Val230 in killifish CYP1A. In silico mutations L387V and V230G enabled TCB to dock closer to the heme in killifish CYP1A. Mutating additional interacting residues (Ala127, Thr233, Asn317, and Tyr386) of killifish CYP1A to the corresponding residues of human CYP1A1 resulted in TCB pose distributions nearly identical with those of human CYP1A1. Docking of TCDD to sets of consensus models of killifish, rat, and human CYP1As showed species differences similar to those with TCB, but with further structural constraints possibly contributing to slower oxidation of TCDD. Docking B[a]P to sets of consensus models of the human and fish CYP1As yielded frequencies of substrate orientations correlating with known regiospecificities for metabolism of B[a]P by these enzymes. The results demonstrate the utility of this ensemble modeling method, which can account for uncertainty inherent in homology modeling and docking by producing statistical distributions of ligand positions.

Performance of the first protein docking server ClusPro in CAPRI rounds 3-5

To evaluate the current status of the protein-protein docking field, the CAPRI experiment came to life. Researchers are given the receptor and ligand 3-dimensional (3D) coordinates before the cocrystallized complex is published. Human predictions of the complex structure are supposed to be submitted within 3 weeks, whereas the server ClusPro has only 24 h and does not make use of any biochemical information. From the 10 targets analyzed in the second evaluation meeting of CAPRI, ClusPro was able to predict meaningful models for 5 targets using only empirical free energy estimates. For two of the targets, the server predictions were assessed to be among the best in the field. Namely, for Targets 8 and 12, ClusPro predicted the model with the most accurate binding-site interface and the model with the highest percentage of nativelike contacts, among 180 and 230 submissions, respectively. After CAPRI, the server has been further developed to predict oligomeric assemblies, and new tools now allow the user to restrict the search for the complex to specific regions on the protein surface, significantly enhancing the predictive capabilities of the server. The performance of ClusPro in CAPRI Rounds 3-5 suggests that clustering the low free energy (i.e., desolvation and electrostatic energy) conformations of a homogeneous conformational sampling of the binding interface is a fast and reliable procedure to detect protein-protein interactions and eliminate false positives. Not including targets that had a significant structural rearrangement upon binding, the success rate of ClusPro was found to be around 71%.

Scoring a diverse set of high-quality docked conformations: A metascore based on electrostatic and desolvation interactions

Predicting protein-protein interactions involves sampling and scoring docked conformations. Barring some large structural rearrangement, rapidly sampling the space of docked conformations is now a real possibility, and the limiting step for the successful prediction of protein interactions is the scoring function used to reduce the space of conformations from billions to a few, and eventually one high affinity complex. An atomic level free-energy scoring function that estimates in units of kcal/mol both electrostatic and desolvation interactions (plus van der Waals if appropriate) of protein-protein docked conformations is used to rerank the blind predictions (860 in total) submitted for six targets to the community-wide Critical Assessment of PRediction of Interactions (CAPRI; http://capri.ebi.ac.uk). We found that native-like models often have varying intermolecular contacts and atom clashes, making unlikely that one can construct a universal function that would rank all these models as native-like. Nevertheless, our scoring function is able to consistently identify the native-like complexes as those with the lowest free energy for the individual models of 16 (out of 17) human predictors for five of the targets, while at the same time the modelers failed to do so in more than half of the cases. The scoring of high-quality models developed by a wide variety of methods and force fields confirms that electrostatic and desolvation forces are the dominant interactions determining the bound structure. The CAPRI experiment has shown that modelers can predict valuable models of protein-protein complexes, and improvements in scoring functions should soon solve the docking problem for complexes whose backbones do not change much upon binding. A scoring server and programs are available at http://structure.pitt.edu.

Optimal clustering for detecting near-native conformations in protein docking

Clustering is one of the most powerful tools in computational biology. The conventional wisdom is that events that occur in clusters are probably not random. In protein docking, the underlying principle is that clustering occurs because long-range electrostatic and/or desolvation forces steer the proteins to a low free-energy attractor at the binding region. Something similar occurs in the docking of small molecules, although in this case shorter-range van der Waals forces play a more critical role. Based on the above, we have developed two different clustering strategies to predict docked conformations based on the clustering properties of a uniform sampling of low free-energy protein-protein and protein-small molecule complexes. We report on significant improvements in the automated prediction and discrimination of docked conformations by using the cluster size and consensus as a ranking criterion. We show that the success of clustering depends on identifying the appropriate clustering radius of the system. The clustering radius for protein-protein complexes is consistent with the range of the electrostatics and desolvation free energies (i.e., between 4 and 9 Angstroms); for protein-small molecule docking, the radius is set by van der Waals interactions (i.e., at approximately 2 Angstroms). Without any a priori information, a simple analysis of the histogram of distance separations between the set of docked conformations can evaluate the clustering properties of the data set. Clustering is observed when the histogram is bimodal. Data clustering is optimal if one chooses the clustering radius to be the minimum after the first peak of the bimodal distribution. We show that using this optimal radius further improves the discrimination of near-native complex structures.

Predicting oligomeric assemblies: N-mers a primer

Multi-protein complexes play key roles in many biological processes. However, since the structures of these assemblies are hard to resolve experimentally, the detailed mechanism of how they work cooperatively in the cell has remained elusive. Similarly, recent advances on in silico prediction of protein-protein interactions have so far avoided this difficult problem. In this paper, we present a general algorithm to predict molecular assemblies of homo-oligomers. Given the number of N-mers and the 3D structure of one monomer, the method samples all the possible symmetries that N-mers can be assembled. Based on a scoring function that clusters the low free energy structures at each binding interface, the algorithm predicts the complex structure as well as the symmetry of the protein assembly. The method is quite general and does not involve any free parameters. The algorithm has been implemented as a public server and integrated to the protein-protein complex prediction server ClusPro. Using this application, we validated predictions for trimers, tetramers (discriminating between dimer of dimers and 4-fold symmetry structures), pentamers and hexamers (discriminating between trimer of dimers, dimer of trimers, and 6-fold symmetry structures), for a total of 107 assemblies. For 85% of the multimers, the server predicts the complex structure within an average rms deviation of 2A from the full crystal. For complexes that involve more than one binding interface, the cluster size at each surface provides a strong indication as to which interface forms first. With improving scoring functions and computer power, our multimer docking approach could be used as a framework to address the more general problem of multi-protein assemblies.

FastContact: rapid estimate of contact and binding free energies

Interaction free energies are crucial for analyzing binding propensities in proteins. Although the problem of computing binding free energies remains open, approximate estimates have become very useful for filtering potential binding complexes. We report on the implementation of a fast computational estimate of the binding free energy based on a statistically determined desolvation contact potential and Coulomb electrostatics with a distance-dependent dielectric constant, and validated in the Critical Assessment of PRotein Interactions experiment. The application also reports residue contact free energies that rapidly highlight the hotspots of the interaction.

AVAILABILITY

The program was written in Fortran. The executable and full documentation is freely available at http://structure.pitt.edu/software/FastContact

Protein-protein docking: is the glass half-full or half-empty?

Are current docking methods capable of building complexes from putative component protein structures? Results of recent computational studies, including those of the CAPRI (Critical Assessment of Protein Interactions) competition, were used to determine the key properties for successful docking and introduce a classification of protein complexes based on docking difficulty. Enzyme-inhibitor complexes could be determined with reasonable accuracy - possibly to within a few alternative structures. Results for antigen-antibody pairs are less predictable, and data for small signaling complexes are generally poor. However, moderate amounts of experimental data can remove uncertainty and the methodology is rapidly improving. Transient complexes with large interface areas undergo substantial conformational change and are beyond the reach of current docking methods. The docking of such complexes might therefore require fundamentally new approaches.

ClusPro: a fully automated algorithm for protein-protein docking

ClusPro (http://nrc.bu.edu/cluster) represents the first fully automated, web-based program for the computational docking of protein structures. Users may upload the coordinate files of two protein structures through ClusPro's web interface, or enter the PDB codes of the respective structures, which ClusPro will then download from the PDB server (http://www.rcsb.org/pdb/). The docking algorithms evaluate billions of putative complexes, retaining a preset number with favorable surface complementarities. A filtering method is then applied to this set of structures, selecting those with good electrostatic and desolvation free energies for further clustering. The program output is a short list of putative complexes ranked according to their clustering properties, which is automatically sent back to the user via email.

Consensus alignment server for reliable comparative modeling with distant templates

Consensus is a server developed to produce high-quality alignments for comparative modeling, and to identify the alignment regions reliable for copying from a given template. This is accomplished even when target-template sequence identity is as low as 5%. Combining the output from five different alignment methods, the server produces a consensus alignment, with a reliability measure indicated for each position and a prediction of the regions suitable for modeling. Models built using the server predictions are typically within 3 A rms deviations from the crystal structure. Users can upload a target protein sequence and specify a template (PDB code); if no template is given, the server will search for one. The method has been validated on a large set of homologous protein structure pairs. The Consensus server should prove useful for modelers for whom the structural reliability of the model is critical in their applications. It is currently available at http://structure.bu.edu/cgi-bin/consensus/consensus.cgi.

Anchor residues in protein-protein interactions

We show that the mechanism for molecular recognition requires one of the interacting proteins, usually the smaller of the two, to anchor a specific side chain in a structurally constrained binding groove of the other protein, providing a steric constraint that helps to stabilize a native-like bound intermediate. We identify the anchor residues in 39 protein-protein complexes and verify that, even in the absence of their interacting partners, the anchor side chains are found in conformations similar to those observed in the bound complex. These ready-made recognition motifs correspond to surface side chains that bury the largest solvent-accessible surface area after forming the complex (> or =100 A2). The existence of such anchors implies that binding pathways can avoid kinetically costly structural rearrangements at the core of the binding interface, allowing for a relatively smooth recognition process. Once anchors are docked, an induced fit process further contributes to forming the final high-affinity complex. This later stage involves flexible (solvent-exposed) side chains that latch to the encounter complex in the periphery of the binding pocket. Our results suggest that the evolutionary conservation of anchor side chains applies to the actual structure that these residues assume before the encounter complex and not just to their loci. Implications for protein docking are also discussed.

ClusPro: an automated docking and discrimination method for the prediction of protein complexes

MOTIVATION

Predicting protein interactions is one of the most challenging problems in functional genomics. Given two proteins known to interact, current docking methods evaluate billions of docked conformations by simple scoring functions, and in addition to near-native structures yield many false positives, i.e. structures with good surface complementarity but far from the native.

RESULTS

We have developed a fast algorithm for filtering docked conformations with good surface complementarity, and ranking them based on their clustering properties. The free energy filters select complexes with lowest desolvation and electrostatic energies. Clustering is then used to smooth the local minima and to select the ones with the broadest energy wells-a property associated with the free energy at the binding site. The robustness of the method was tested on sets of 2000 docked conformations generated for 48 pairs of interacting proteins. In 31 of these cases, the top 10 predictions include at least one near-native complex, with an average RMSD of 5 A from the native structure. The docking and discrimination method also provides good results for a number of complexes that were used as targets in the Critical Assessment of PRedictions of Interactions experiment.

AVAILABILITY

The fully automated docking and discrimination server ClusPro can be found at http://structure.bu.edu

Consensus alignment for reliable framework prediction in homology modeling

MOTIVATION

Even the best sequence alignment methods frequently fail to correctly identify the framework regions for which backbones can be copied from the template into the target structure. Since the underprediction and, more significantly, the overprediction of these regions reduces the quality of the final model, it is of prime importance to attain as much as possible of the true structural alignment between target and template.

RESULTS

We have developed an algorithm called Consensus that consistently provides a high quality alignment for comparative modeling. The method follows from a benchmark analysis of the 3D models generated by ten alignment techniques for a set of 79 homologous protein structure pairs. For 20-to-40% of the targets, these methods yield models with at least 6 A root mean square deviation (RMSD) from the native structure. We have selected the top five performing methods, and developed a consensus algorithm to generate an improved alignment. By building on the individual strength of each method, a set of criteria was implemented to remove the alignment segments that are likely to correspond to structurally dissimilar regions. The automated algorithm was validated on a different set of 48 protein pairs, resulting in 2.2 A average RMSD for the predicted models, and only four cases in which the RMSD exceeded 3 A. The average length of the alignments was about 75% of that found by standard structural superposition methods. The performance of Consensus was consistent from 2 to 32% target-template sequence identity, and hence it can be used for accurate prediction of framework regions in homology modeling.

Successful discrimination of protein interactions

We present results from the prediction of protein complexes associated with the first Critical Assessment of PRediction of Interactions (CAPRI) experiment. Our algorithm, SmoothDock, comprises four steps: (1) we perform rigid body docking using the program DOT, keeping the top 20,000 structures as ranked by surface complementarity; (2) we rerank these structures according to a free energy estimate that includes both desolvation and electrostatics and retain the top 2000 complexes; (3) we cluster the filtered complexes using a pairwise root-mean-square deviation (RMSD) criterion; (4) the 25 largest clusters are subject to a smooth docking discrimination algorithm where van der Waals forces are taken into account. We predicted targets 1, 6, and 7 with RMSDs of 9.5, 2.4, and 2.6 A, respectively. More importantly, from the perspective of biological applications, our approach consistently ranked the correct model first (i.e., with highest confidence). For target 5 we identified the binding region but not the correct orientation. Although we were able to find reasonable clusters for all targets, low-affinity complexes (K(d) < nM) were harder to discriminate. For four of seven targets, the top models predicted by our automated procedure were among the best communitywide predictions.

Protein-protein association kinetics and protein docking

Rigid body protein docking methods frequently yield false positive structures that have good surface complementarity, but are far from the native complex. The main reason for this is the uncertainty of the protein structures to be docked, including the positions of solvent-exposed sidechains. Substantial efforts have been devoted to finding near-native structures by rescoring the docked conformations and employing various filters. An alternative approach emulates the process of protein-protein association, that is, first finding the region in which binding is likely to occur and then refining the complex while allowing for flexibility.

Protein docking along smooth association pathways

We propose a docking method that mimics the way proteins bind. The method accounts for the dominant driving forces at the different length scales of the protein binding process, allowing for an efficient selection of a downhill path on the evolving receptor-ligand-free energy landscape. Starting from encounter complexes with as much as 10 A rms deviation from the native conformation, the method locally samples the six dimensional space of rigid-body receptor-ligand structures subject to a van der Waals constraint. The sampling is initially biased only by the desolvation and electrostatic components of the free energy, which capture the partial affinity of unbound structures that are more than 4 A away from the native state. Below this threshold, improved discrimination is attained by adding an increasing fraction of the van der Waals energy to the force field. The method, with no free parameters, was tested in eight different sets of independently crystallized receptor-ligand structures consistently predicting bound conformations with the lowest free energies and appropriate stability gap around 2 A from the native complex. This multistage approach is consistent with the underlying kinetics and internal structure of the free energy funnel to the bound state. Implications for the nature of the protein binding pathways are also discussed.

Dynamical view of the positions of key side chains in protein-protein recognition

When a complex is constructed from the separately determined rigid structures of a receptor and its ligand, some key side chains are usually in wrong positions. These distortions of the interface yield an apparent loss in affinity and would unfavorably affect the kinetics of association. It is generally assumed that the interacting proteins should drive the appropriate conformational changes, leading to their complementarity, but this hypothesis does not explain their fast association rates. However, nanosecond explicit solvent molecular dynamics simulations of misfolded surface side chains from the independently solved structures of barstar, bovine pancreatic trypsin inhibitor, and lysozyme show that even before any receptor-ligand interaction, key side chains frequently visit the rotamer conformations seen in the complex. We show that these simple structural motifs can reconcile most of the binding affinity required for a rapid and highly specific association process. Side chains amenable to induced fit are also identified. These results corroborate that solvent-side chain interactions play a critical role in the recognition process. Our findings are also supported by crystallographic data.

Predictions of gene family distributions in microbial genomes: evolution by gene duplication and modification

A universal property of microbial genomes is the considerable fraction of genes that are homologous to other genes within the same genome. The process by which these homologues are generated is not well understood, but sequence analysis of 20 microbial genomes unveils a recurrent distribution of gene family sizes. We show that a simple evolutionary model based on random gene duplication and point mutations fully accounts for these distributions and permits predictions for the number of gene families in genomes not yet complete. Our findings are consistent with the notion that a genome evolves from a set of precursor genes to a mature size by gene duplications and increasing modifications.

Scoring docked conformations generated by rigid-body protein-protein docking

Rigid-body methods, particularly Fourier correlation techniques, are very efficient for docking bound (co-crystallized) protein conformations using measures of surface complementarity as the target function. However, when docking unbound (separately crystallized) conformations, the method generally yields hundreds of false positive structures with good scores but high root mean square deviations (RMSDs). This paper describes a two-step scoring algorithm that can discriminate near-native conformations (with less than 5 A RMSD) from other structures. The first step includes two rigid-body filters that use the desolvation free energy and the electrostatic energy to select a manageable number of conformations for further processing, but are unable to eliminate all false positives. Complete discrimination is achieved in the second step that minimizes the molecular mechanics energy of the retained structures, and re-ranks them with a combined free-energy function which includes electrostatic, solvation, and van der Waals energy terms. After minimization, the improved fit in near-native complex conformations provides the free-energy gap required for discrimination. The algorithm has been developed and tested using docking decoys, i.e., docked conformations generated by Fourier correlation techniques. The decoy sets are available on the web for testing other discrimination procedures. Proteins 2000;40:525-537.

Kinetics of desolvation-mediated protein-protein binding

The role of desolvation in protein binding kinetics is investigated using Brownian dynamics simulations in complexes in which the electrostatic interactions are relatively weak. We find that partial desolvation, modeled by a short-range atomic contact potential, is not only a major contributor to the binding free energy but also substantially increases the diffusion-limited rate for complexes in which long-range electrostatics is weak. This rate enhancement is mostly due to weakly specific pathways leading to a low free-energy attractor, i.e., a precursor state before docking. For alpha-chymotrypsin and human leukocyte elastase, both interacting with turkey ovomucoid third domain, we find that the forward rate constant associated with a collision within a solid angle phi around their corresponding attractor approaches 10(7) and 10(6) M(-1)s(-1), respectively, in the limit phi approximately 2 degrees. Because these estimates agree well with experiments, we conclude that the final bound conformation must be preceded by a small set of well-defined diffusion-accessible precursor states. The inclusion of the otherwise repulsive desolvation interaction also explains the lack of aggregation in proteins by restricting nonspecific association times to approximately 4 ns. Under the same reaction conditions but without short range forces, the association rate would be only approximately 10(3) M(-1)s(-1). Although desolvation increases these rates by three orders of magnitude, desolvation-mediated association is still at least 100-fold slower than the electrostatically assisted binding in complexes such as barnase and barstar.

Promotion and prevention choices between stability and change

Two situations involving choice between stability and change were examined: task substitution, which deals with choosing between resuming an interrupted activity and doing a substitute activity, and endowment, which deals with choosing between a possessed object and an alternative object. Regulatory focus theory (E. T. Higgins, 1997, 1998) predicts that a promotion focus will be associated with openness to change, whereas a prevention focus will be associated with a preference for stability. Five studies confirmed this prediction with both situational induction of and chronic personality differences in regulatory focus. In Studies 1 and 2, individuals in a prevention focus were more inclined than individuals in a promotion focus to resume an interrupted task rather than do a substitute task. In Studies 3-5, individuals in a prevention focus, but not individuals in a promotion focus, exhibited a reluctance to exchange currently possessed objects (i.e., endowment) or previously possessed objects.

Continuum electrostatic analysis of preferred solvation sites around proteins in solution

To understand water-protein interactions in solution, the electrostatic field is calculated by solving the Poisson-Boltzmann equation, and the free energy surface of water is mapped by translating and rotating an explicit water molecule around the protein. The calculation is applied to T4 lysozyme with data available on the conservation of solvent binding sites in 18 crystallographically independent molecules. The free energy maps around the ordered water sites provide information on the relationship between water positions in crystal structure and in solution. Results show that almost all conserved sites and the majority of nonconserved sites are within 1.3 A of local free energy minima. This finding is in sharp contrast to the behavior of randomly placed water molecules in the boundary layer, which, on the average, must travel more than 3 A to the nearest free energy minimum. Thus, the solvation sites are at least partially determined by protein-water interactions rather than by crystal packing alone. The characteristic water residence times, obtained from the free energies at the local minima, are in good agreement with nuclear magnetic resonance experiments. Only about half of the potential sites show up as ordered water in the 1.7 A resolution X-ray structure. Crystal packing interactions can stabilize weak or mobile potential sites (in fact, some ordered water positions are not close to free energy minima) or can prevent water from occupying certain sites. Apart from a few buried water molecules that are strong binders, the free energies are not very different for conserved and nonconserved sites. We show that conservation of a water site between two crystals occurs if the positions of protein atoms, primarily contributing to the free energy at the local minimum, do not substantially change from one structure to the other. This requirement can be correlated with the nature of the side chain contacting the water molecule in the site.

Promotion and prevention choices between stability and change

Two situations involving choice between stability and change were examined: task substitution, which deals with choosing between resuming an interrupted activity and doing a substitute activity, and endowment, which deals with choosing between a possessed object and an alternative object. Regulatory focus theory (E. T. Higgins, 1997, 1998) predicts that a promotion focus will be associated with openness to change, whereas a prevention focus will be associated with a preference for stability. Five studies confirmed this prediction with both situational induction of and chronic personality differences in regulatory focus. In Studies 1 and 2, individuals in a prevention focus were more inclined than individuals in a promotion focus to resume an interrupted task rather than do a substitute task. In Studies 3-5, individuals in a prevention focus, but not individuals in a promotion focus, exhibited a reluctance to exchange currently possessed objects (i.e., endowment) or previously possessed objects.

Free energy landscapes of encounter complexes in protein-protein association

We report the computer generation of a high-density map of the thermodynamic properties of the diffusion-accessible encounter conformations of four receptor-ligand protein pairs, and use it to study the electrostatic and desolvation components of the free energy of association. Encounter complex conformations are generated by sampling the translational/rotational space of the ligand around the receptor, both at 5-A and zero surface-to-surface separations. We find that partial desolvation is always an important effect, and it becomes dominant for complexes in which one of the reactants is neutral or weakly charged. The interaction provides a slowly varying attractive force over a small but significant region of the molecular surface. In complexes with no strong charge complementarity this region surrounds the binding site, and the orientation of the ligand in the encounter conformation with the lowest desolvation free energy is similar to the one observed in the fully formed complex. Complexes with strong opposite charges exhibit two types of behavior. In the first group, represented by barnase/barstar, electrostatics exerts strong orientational steering toward the binding site, and desolvation provides some added adhesion within the local region of low electrostatic energy. In the second group, represented by the complex of kallikrein and pancreatic trypsin inhibitor, the overall stability results from the rather nonspecific electrostatic attraction, whereas the affinity toward the binding region is determined by desolvation interactions.

Denaturants can accelerate folding rates in a class of globular proteins

We present a lattice Monte Carlo study to examine the effect of denaturants on the folding rates of simplified models of proteins. The two-dimensional model is made from a three-letter code mimicking the presence of hydrophobic, hydrophilic, and cysteine residues. We show that the rate of folding is maximum when the effective hydrophobic interaction epsilon H is approximately equal to the free energy gain epsilon S upon forming disulfide bonds. In the range 1 < or = epsilon H/ epsilon S < or = 3, multiple paths that connect several intermediates to the native state lead to fast folding. It is shown that at a fixed temperature and epsilon S the folding rate increases as epsilon H decreases. An approximate model is used to show that epsilon H should decrease as a function of the concentration of denaturants such as urea or guanidine hydrochloride. Our simulation results, in conjunction with this model, are used to show that increasing the concentration of denaturants can lead to an increase in folding rates. This occurs because denaturants can destabilize the intermediates without significantly altering the energy of the native conformation. Our findings are compared with experiments on the effects of denaturants on the refolding of bovine pancreatic trypsin inhibitor and ribonuclease T1. We also argue that the phenomenon of denaturant-enhanced folding of proteins should be general.

Modeling the role of disulfide bonds in protein folding: entropic barriers and pathways

The role of disulfide bonds in directing protein folding is studied using lattice models. We find that the stability and the specificity of the disulfide bond interactions play quite different roles in the folding process: Under some conditions, the stability decreases the overall rate of folding; the specificity, however, by yielding a simpler connectivity of intermediates, always increases the rate of folding. This conclusion is intimately related to the selection mechanism entailed by entropic driving forces, such as the loop formation probability, and entropic barriers separating the native and the many native-like metastable states. The folding time is found to be a minimum for a certain range of the effective disulfide bond interaction. Examination of a model, which allows for the formation of disulfide bonded intermediates, suggests that folding proceeds via a three-stage multiple pathways kinetics. We show that there are pathways to the native state involving only native-like intermediates, as well as those that are mediated by nonnative intermediates. These findings are interpreted in terms of the appropriate energy landscape describing the barriers connecting low energy conformations. The consistency of our conclusions with several experimental studies is also discussed.

Theoretical predictions of folding pathways by using the proximity rule, with applications to bovine pancreatic trypsin inhibitor

We propose a phenomenological theory that accounts for entropic effects due to loop formation to predict pathways in the kinetics of protein folding. The theory, the basis of which lies in multiple folding pathways and a three-stage kinetics, qualitatively reproduces most of the kinetic measurements in the refolding of bovine pancreatic trypsin inhibitor. The resulting pathways show that nonnative kinetic transients are involved in the productive routes leading to the formation of native intermediates. Our theory emphasizes the importance of the random origin of chain folding initiation structures in directing protein folding.

Kinetics and thermodynamics of folding in model proteins

Monte Carlo simulations on a class of lattice models are used to probe the thermodynamics and kinetics of protein folding. We find two transition temperatures: one at T theta, when chains collapse from a coil to a compact phase, and the other at Tf (< T theta), when chains adopt a conformation corresponding to their native state. The kinetics are probed by several correlation functions and are interpreted in terms of the underlying energy landscape. The transition from the coil to the native state occurs in three distinct stages. The initial stage corresponds to a random collapse of the protein chain. At intermediate times tau c, during which much of the native structure is acquired, there are multiple pathways. For longer times tau r (>> tau c) the decay is exponential, suggestive of a late transition state. The folding time scale (approximately tau r) varies greatly depending on the model. Implications of our results for in vitro folding of proteins are discussed.