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

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

Characterization of low affinity Fcγ receptor biotinylation under controlled reaction conditions by mass spectrometry and ligand binding analysis

Chemical protein biotinylation and streptavidin or anti‐biotin‐based capture is regularly used for proteins as a more controlled alternative to direct coupling of the protein on a biosensor surface. On biotinylation an interaction site of interest may be blocked by the biotin groups, diminishing apparent activity of the protein. Minimal biotinylation can circumvent the loss of apparent activity, but still a binding site of interest can be blocked when labeling an amino acid involved in the binding. Here, we describe reaction condition optimization studies for minimal labeling. We have chosen low affinity Fcγ receptors as model compounds as these proteins contain many lysines in their active binding site and as such provide an interesting system for a minimal labeling approach. We were able to identify the most critical parameters (protein:biotin ratio and incubation pH) for a minimal labeling approach in which the proteins of choice remain most active toward analyte binding. Localization of biotinylation by mass spectrometric peptide mapping on minimally labeled material was correlated to protein activity in binding assays. We show that only aiming at minimal labeling is not sufficient to maintain an active protein. Careful fine‐tuning of critical parameters is important to reduce biotinylation in a protein binding site.     

Mapping discontinuous protein‐binding sites via structure‐based peptide libraries: combining in silico and in vitro approaches

To perform their various functions, protein surfaces often have to interact with each other in a specific way. Usually, only parts of a protein are accessible and can act as binding sites. Because proteins consist of polypeptide chains that fold into complex three‐dimensional shapes, binding sites can be divided into two different types: linear sites that follow the primary amino acid sequence and discontinuous binding sites, which are made up of short peptide fragments that are adjacent in spatial proximity. Such discontinuous binding sites dominate protein–protein interactions, but are difficult to identify. To meet this challenge, we combined a computational, structure‐based approach and an experimental, high‐throughput method. SUPERFICIAL is a program that uses protein structures as input and generates peptide libraries to represent the protein's surface. A large number of the predicted peptides can be simultaneously synthesised applying the SPOT technology. The results of a binding assay subsequently help to elucidate protein–protein interactions; the approach is applicable to any kind of protein. The crystal structure of the complex of hen egg lysozyme with the well‐characterised murine IgG1 antibody HyHEL‐5 is available, and the complex is known to have a discontinuous binding site. Using SUPERFICIAL, the entire surface of lysozyme was translated into a peptide library that was synthesised on a cellulose membrane using the SPOT technology and tested against the HyHEL‐5 antibody. In this way, it was possible to identify two peptides (longest common sequence and peptide 19) that represented the discontinuous epitope of lysozyme. Copyright © 2012 John Wiley & Sons, Ltd.     

OptGraft: A computational procedure for transferring a binding site onto an existing protein scaffold

One of the many challenging tasks of protein design is the introduction of a completely new function into an existing protein scaffold. In this study, we introduce a new computational procedure OptGraft for placing a novel binding pocket onto a protein structure so as its geometry is minimally perturbed. This is accomplished by introducing a two‐level procedure where we first identify where are the most appropriate locations to graft the new binding pocket into the protein fold by minimizing the departure from a set of geometric restraints using mixed‐integer linear optimization. On identifying the suitable locations that can accommodate the new binding pocket, CHARMM energy calculations are employed to identify what mutations in the neighboring residues, if any, are needed to ensure that the minimum energy conformation of the binding pocket conserves the desired geometry. This computational framework is benchmarked against the results available in the literature for engineering a copper binding site into thioredoxin protein. Subsequently, OptGraft is used to guide the transfer of a calcium‐binding pocket from thermitase protein (PDB: 1thm) into the first domain of CD2 protein (PDB:1hng). Experimental characterization of three de novo redesigned proteins with grafted calcium‐binding centers demonstrated that they all exhibit high affinities for terbium (K_d ～ 22, 38, and 55 μM) and can selectively bind calcium over magnesium.     

DynaBiS: A hierarchical sampling algorithm to identify flexible binding sites for large ligands and peptides

Knowing the ligand or peptide binding site in proteins is highly important to guide drug discovery, but experimental elucidation of the binding site is difficult. Therefore, various computational approaches have been developed to identify potential binding sites in protein structures. However, protein and ligand flexibility are often neglected in these methods due to efficiency considerations despite the recognition that protein–ligand interactions can be strongly affected by mutual structural adaptations. This is particularly true if the binding site is unknown, as the screening will typically be performed based on an unbound protein structure. Herein we present DynaBiS, a hierarchical sampling algorithm to identify flexible binding sites for a target ligand with explicit consideration of protein and ligand flexibility, inspired by our previously presented flexible docking algorithm DynaDock. DynaBiS applies soft-core potentials between the ligand and the protein, thereby allowing a certain protein–ligand overlap resulting in efficient sampling of conformational adaptation effects. We evaluated DynaBiS and other commonly used binding site identification algorithms against a diverse evaluation set consisting of 26 proteins featuring peptide as well as small ligand binding sites. We show that DynaBiS outperforms the other evaluated methods for the identification of protein binding sites for large and highly flexible ligands such as peptides, both with a holo or apo structure used as input.     

Target Detection Assay (TDA): a versatile procedure to determine DNA binding sites as demonstrated on SP1 protein.

We developed a rapid method designated Target Detection Assay (TDA) to determine DNA binding sites for putative DNA binding proteins. A purified, functionally active DNA binding protein and a pool of random double-stranded oligonucleotides harbouring PCR primer sites at each end are included the TDA cycle which consists of four separate steps: a DNA protein incubation step, a protein DNA complex separation step, a DNA elution step and a polymerase chain reaction (PCR) DNA amplification step. The stringency of selection can be increased in consecutive TDA cycles. Since tiny amounts of retained DNA can be rescued by PCR, buffer systems, salt concentrations and competitor DNA contents can be varied in order to determine high affinity binding sites for the protein of choice. To test the efficiency of the TDA procedure potential DNA binding sites were selected by the DNA binding protein SP1 from a pool of oligonucleotides with random nucleotides at 12 positions. Target sites selected by recombinant SP1 closely matched the SP1 consensus site. If DNA recognition sites have to be determined for known, mutated or putative DNA binding proteins, the Target Detection Assay (TDA) is a versatile and rapid technique for consideration.     

Efficient identification of side-chain patterns using a multidimensional index tree

Convergent evolution often produces similar functional sites in nonhomologous proteins. The identification of these sites can make it possible to infer function from structure, to pinpoint the location of a functional site, to identify enzymes with similar enzymatic mechanisms, or to discover putative functional sites. In this article, a novel method is presented that (a) queries a database of protein structures for the occurrence of a given side chain pattern and (b) identifies interesting side-chain patterns in a given structure. For efficiency and to make a robust statistical evaluation of the significance of a similarity possible, patterns of three residues (or triads) are considered. Each triad is encoded as a high-dimensional vector and stored in an SR (Sphere/Rectangle) tree, an efficient multidimensional index tree. Identifying similar triads can then be reformulated as identifying neighboring vectors. The method deals with many features that otherwise complicate the identification of meaningful patterns: shifted backbone positions, conservative substitutions, various atom label ambiguities and mirror imaged geometries. The combined treatment of these features leads to the identification of previously unidentified patterns. In particular, the identification of mirror imaged side-chain patterns is unique to the here-described method. Interesting triads in a given structure can be identified by extracting all triads and comparing them with a database of triads involved in ligand binding. The approach was tested by an all-against-all comparison of unique representatives of all SCOP superfamilies. New findings include mirror imaged metal binding and active sites, and a putative active site in bacterial luciferase.     

BiPPred: Combined sequence‐ and structure‐based prediction of peptide binding to the Hsp70 chaperone BiP

Substrate binding to Hsp70 chaperones is involved in many biological processes, and the identification of potential substrates is important for a comprehensive understanding of these events. We present a multi‐scale pipeline for an accurate, yet efficient prediction of peptides binding to the Hsp70 chaperone BiP by combining sequence‐based prediction with molecular docking and MMPBSA calculations. First, we measured the binding of 15mer peptides from known substrate proteins of BiP by peptide array (PA) experiments and performed an accuracy assessment of the PA data by fluorescence anisotropy studies. Several sequence‐based prediction models were fitted using this and other peptide binding data. A structure‐based position‐specific scoring matrix (SB‐PSSM) derived solely from structural modeling data forms the core of all models. The matrix elements are based on a combination of binding energy estimations, molecular dynamics simulations, and analysis of the BiP binding site, which led to new insights into the peptide binding specificities of the chaperone. Using this SB‐PSSM, peptide binders could be predicted with high selectivity even without training of the model on experimental data. Additional training further increased the prediction accuracies. Subsequent molecular docking (DynaDock) and MMGBSA/MMPBSA‐based binding affinity estimations for predicted binders allowed the identification of the correct binding mode of the peptides as well as the calculation of nearly quantitative binding affinities. The general concept behind the developed multi‐scale pipeline can readily be applied to other protein‐peptide complexes with linearly bound peptides, for which sufficient experimental binding data for the training of classical sequence‐based prediction models is not available. Proteins 2016; 84:1390–1407. © 2016 Wiley Periodicals, Inc.     

Comparison analysis of primary ligand‐binding sites in seven‐helix membrane proteins

Seven‐helix transmembrane proteins, including the G‐protein‐coupled receptors (GPCRs), mediate a broad range of fundamental cellular activities through binding to a wide range of ligands. Understanding the structural basis for the ligand‐binding selectivity of these proteins is of significance to their structure‐based drug design. Comparison analysis of proteins' ligand‐binding sites provides a useful way to study their structure‐activity relationships. Various computational methods have been developed for the binding‐site comparison of soluble proteins. In this work, we applied this approach to the analysis of the primary ligand‐binding sites of 92 seven‐helix transmembrane proteins. Results of the studies confirmed that the binding site of bacterial rhodopsins is indeed different from all GPCRs. In the latter group, further comparison of the binding sites indicated a group of residues that could be responsible for ligand‐binding selectivity and important for structure‐based drug design. Furthermore, unexpected binding‐site dissimilarities were observed among adrenergic and adenosine receptors, suggesting that the percentage of the overall sequence identity between a target protein and a template protein alone is not sufficient for selecting the best template for homology modeling of seven‐helix membrane proteins. These results provided novel insight into the structural basis of ligand‐binding selectivity of seven‐helix membrane proteins and are of practical use to the computational modeling of these proteins. © 2010 Wiley Periodicals, Inc. Biopolymers 95: 31–38, 2011.     

Allosteric sites can be identified based on the residue–residue interaction energy difference

Allosteric drugs act at a distance to regulate protein functions. They have several advantages over conventional orthosteric drugs, including diverse regulation types and fewer side effects. However, the rational design of allosteric ligands remains a challenge, especially when it comes to the identification allosteric binding sites. As the binding of allosteric ligands may induce changes in the pattern of residue–residue interactions, we calculated the residue–residue interaction energies within the allosteric site based on the molecular mechanics generalized Born surface area energy decomposition scheme. Using a dataset of 17 allosteric proteins with structural data for both the apo and the ligand‐bound state available, we used conformational ensembles generated by molecular dynamics simulations to compute the differences in the residue–residue interaction energies in known allosteric sites from both states. For all the known sites, distinct interaction energy differences (>25%) were observed. We then used CAVITY, a binding site detection program to identify novel putative allosteric sites in the same proteins. This yielded a total of 31 “druggable binding sites,” of which 21 exhibited >25% difference in residue interaction energies, and were hence predicted as novel allosteric sites. Three of the predicted allosteric sites were supported by recent experimental studies. All the predicted sites may serve as novel allosteric sites for allosteric ligand design. Our study provides a computational method for identifying novel allosteric sites for allosteric drug design. Proteins 2015; 83:1375–1384. © 2014 Wiley Periodicals, Inc.     

Three enhancements to the inference of statistical protein‐DNA potentials

The energetics of protein‐DNA interactions are often modeled using so‐called statistical potentials, that is, energy models derived from the atomic structures of protein‐DNA complexes. Many statistical protein‐DNA potentials based on differing theoretical assumptions have been investigated, but little attention has been paid to the types of data and the parameter estimation process used in deriving the statistical potentials. We describe three enhancements to statistical potential inference that significantly improve the accuracy of predicted protein‐DNA interactions: (i) incorporation of binding energy data of protein‐DNA complexes, in conjunction with their X‐ray crystal structures, (ii) use of spatially‐aware parameter fitting, and (iii) use of ensemble‐based parameter fitting. We apply these enhancements to three widely‐used statistical potentials and use the resulting enhanced potentials in a structure‐based prediction of the DNA binding sites of proteins. These enhancements are directly applicable to all statistical potentials used in protein‐DNA modeling, and we show that they can improve the accuracy of predicted DNA binding sites by up to 21%. Proteins 2013. © 2012 Wiley Periodicals, Inc.     

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

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

Prediction of protein‐glucose binding sites using support vector machines

Glucose is a simple sugar that plays an essential role in many basic metabolic and signaling pathways. Many proteins have binding sites that are highly specific to glucose. The exponential increase of genomic data has revealed the identity of many proteins that seem to be central to biological processes, but whose exact functions are unknown. Many of these proteins seem to be associated with disease processes. Being able to predict glucose‐specific binding sites in these proteins will greatly enhance our ability to annotate protein function and may significantly contribute to drug design. We hereby present the first glucose‐binding site classifier algorithm. We consider the sugar‐binding pocket as a spherical spatio‐chemical environment and represent it as a vector of geometric and chemical features. We then perform Random Forests feature selection to identify key features and analyze them using support vector machines classification. Our work shows that glucose binding sites can be modeled effectively using a limited number of basic chemical and residue features. Using a leave‐one‐out cross‐validation method, our classifier achieves a 8.11% error, a 89.66% sensitivity and a 93.33% specificity over our dataset. From a biochemical perspective, our results support the relevance of ordered water molecules and ions in determining glucose specificity. They also reveal the importance of carboxylate residues in glucose binding and the high concentration of negatively charged atoms in direct contact with the bound glucose molecule. Proteins 2009. © 2009 Wiley‐Liss, Inc.     

Interaction of serum amyloid A with human cystatin C—identification of binding sites

Serum amyloid A (SAA) is a multifunctional acute‐phase protein whose natural role seems to be participation in many physiologic and pathological processes. Prolonged increased SAA level in a number of chronic inflammatory and neoplastic diseases gives rise to reactive systemic amyloid A amyloidosis, where the N‐terminal 76‐amino acid residue‐long segment of SAA is deposited as amyloid fibrils. Recently, a specific interaction between SAA and the ubiquitous inhibitor of cysteine proteases—human cystatin C (hCC)—has been described. Here, we report further evidence corroborating this interaction, and the identification of the SAA and hCC binding sites in the SAA–hCC complex, using a combination of selective proteolytic excision and high‐resolution mass spectrometry. The shortest binding site in the SAA sequence was determined as SAA(86–104), whereas the binding site in hCC sequence was identified as hCC(96–102). Binding specificities of both interacting sequences were ascertained by affinity experiments (ELISA) and by registration of mass spectrum of SAA–hCC complex. Copyright © 2012 John Wiley & Sons, Ltd.     

Single‐step protease cleavage elution for identification of protein–protein interactions from GST pull‐down and mass spectrometry

The study of protein–protein interactions is a major theme in biological disciplines. Pull‐down or affinity‐precipitation assays using GST fusion proteins have become one of the most common and valuable approaches to identify novel binding partners for proteins of interest (bait). Non‐specific binding of prey proteins to the beads or to GST itself, however, inevitably complicates and impedes subsequent analysis of pull‐down results. A variety of measures, each with inherent advantages and limitations, can minimise the extent of the background. This technical brief details and tests a modification of established GST pull‐down protocols. By specifically eluting only the bait (minus the GST tag) and the associated non‐specific binding proteins with a simple, single‐step protease cleavage, a cleaner platform for downstream protein identification with MS is established. We present a proof of concept for this method, as evidenced by a GST pull‐down/MS case study of the small guanosine triphosphatase (GTPase) Rab31 in which: (i) sensitivity was enhanced, (ii) a reduced level of background was observed, (iii) distinguishability of non‐specific contaminant proteins from genuine binders was improved and (iv) a putative new protein–protein interaction was discovered. Our protease cleavage step is readily applicable to all further affinity tag pull‐downs.