A scoring function for docking ligands to low-resolution protein structures

Protein surface roughness accounts for binding free energy of Plasmepsin II-ligand complexes

The calculation of absolute binding affinities for protein-inhibitor complexes remains as one of the main challenges in computational structure-based ligand design. The present work explored the calculations of surface fractal dimension (as a measure of surface roughness) and the relationship with experimental binding free energies of Plasmepsin II complexes. Plasmepsin II is an attractive target for novel therapeutic compounds to treat malaria. However, the structural flexibility of this enzyme is a drawback when searching for specific inhibitors. Concerning that, we performed separate explicitly solvated molecular dynamics simulations using the available high-resolution crystal structures of different Plasmepsin II complexes. Molecular dynamics simulations allowed a better approximation to systems dynamics and, therefore, a more reliable estimation of surface roughness. This constitutes a novel approximation in order to obtain more realistic values of fractal dimension, because previous works considered only x-ray structures. Binding site fractal dimension was calculated considering the ensemble of structures generated at different simulation times. A linear relationship between binding site fractal dimension and experimental binding free energies of the complexes was observed within 20 ns. Previous studies of the subject did not uncover this relationship. Regression model, coined FD model, was built to estimate binding free energies from binding site fractal dimension values. Leave-one-out cross-validation showed that our model reproduced accurately the absolute binding free energies for our training set (R²  = 0.76; <|error|> =0.55 kcal/mol; SD_error  = 0.19 kcal/mol). The fact that such a simple model may be applied raises some questions that are addressed in the article.

eMatchSite: sequence order-independent structure alignments of ligand binding pockets in protein models

Detecting similarities between ligand binding sites in the absence of global homology between target proteins has been recognized as one of the critical components of modern drug discovery. Local binding site alignments can be constructed using sequence order-independent techniques, however, to achieve a high accuracy, many current algorithms for binding site comparison require high-quality experimental protein structures, preferably in the bound conformational state. This, in turn, complicates proteome scale applications, where only various quality structure models are available for the majority of gene products. To improve the state-of-the-art, we developed eMatchSite, a new method for constructing sequence order-independent alignments of ligand binding sites in protein models. Large-scale benchmarking calculations using adenine-binding pockets in crystal structures demonstrate that eMatchSite generates accurate alignments for almost three times more protein pairs than SOIPPA. More importantly, eMatchSite offers a high tolerance to structural distortions in ligand binding regions in protein models. For example, the percentage of correctly aligned pairs of adenine-binding sites in weakly homologous protein models is only 4–9% lower than those aligned using crystal structures. This represents a significant improvement over other algorithms, e.g. the performance of eMatchSite in recognizing similar binding sites is 6% and 13% higher than that of SiteEngine using high- and moderate-quality protein models, respectively. Constructing biologically correct alignments using predicted ligand binding sites in protein models opens up the possibility to investigate drug-protein interaction networks for complete proteomes with prospective systems-level applications in polypharmacology and rational drug repositioning. eMatchSite is freely available to the academic community as a web-server and a stand-alone software distribution at http://www.brylinski.org/ematchsite.

eFindSite: Enhanced Fingerprint-Based Virtual Screening Against Predicted Ligand Binding Sites in Protein Models

A standard practice for lead identification in drug discovery is ligand virtual screening, which utilizes computing technologies to detect small compounds that likely bind to target proteins prior to experimental screens. A high accuracy is often achieved when the target protein has a resolved crystal structure; however, using protein models still renders significant challenges. Towards this goal, we recently developed eFindSite that predicts ligand binding sites using a collection of effective algorithms, including meta-threading, machine learning and reliable confidence estimation systems. Here, we incorporate fingerprint-based virtual screening capabilities in eFindSite in addition to its flagship role as a ligand binding pocket predictor. Virtual screening benchmarks using the enhanced Directory of Useful Decoys demonstrate that eFindSite significantly outperforms AutoDock Vina as assessed by several evaluation metrics. Importantly, this holds true regardless of the quality of target protein structures. As a first genome-wide application of eFindSite, we conduct large-scale virtual screening of the entire proteome of Escherichia coli with encouraging results. In the new approach to fingerprint-based virtual screening using remote protein homology, eFindSite demonstrates its compelling proficiency offering a high ranking accuracy and low susceptibility to target structure deformations. The enhanced version of eFindSite is freely available to the academic community at http://www.brylinski.org/efindsite.

eFindSite: improved prediction of ligand binding sites in protein models using meta-threading, machine learning and auxiliary ligands

Molecular structures and functions of the majority of proteins across different species are yet to be identified. Much needed functional annotation of these gene products often benefits from the knowledge of protein-ligand interactions. Towards this goal, we developed eFindSite, an improved version of FINDSITE, designed to more efficiently identify ligand binding sites and residues using only weakly homologous templates. It employs a collection of effective algorithms, including highly sensitive meta-threading approaches, improved clustering techniques, advanced machine learning methods and reliable confidence estimation systems. Depending on the quality of target protein structures, eFindSite outperforms geometric pocket detection algorithms by 15-40 % in binding site detection and by 5-35 % in binding residue prediction. Moreover, compared to FINDSITE, it identifies 14 % more binding residues in the most difficult cases. When multiple putative binding pockets are identified, the ranking accuracy is 75-78 %, which can be further improved by 3-4 % by including auxiliary information on binding ligands extracted from biomedical literature. As a first across-genome application, we describe structure modeling and binding site prediction for the entire proteome of Escherichia coli. Carefully calibrated confidence estimates strongly indicate that highly reliable ligand binding predictions are made for the majority of gene products, thus eFindSite holds a significant promise for large-scale genome annotation and drug development projects. eFindSite is freely available to the academic community at http://www.brylinski.org/efindsite .

Novel insights of structure-based modeling for RNA-targeted drug discovery

Substantial progress in RNA biology highlights the importance of RNAs (e.g., microRNAs) in diseases and the potential of targeting RNAs for drug discovery. However, the lack of RNA-specific modeling techniques demands the development of new tools for RNA-targeted rational drug design. Herein, we implemented integrated approaches of accurate RNA modeling and virtual screening for RNA inhibitor discovery with the most comprehensive evaluation to date of five docking and 11 scoring methods. For the first time, statistical analysis was heavily employed to assess the significance of our predictions. We found that GOLD:GOLD Fitness and rDock:rDock_solv could accurately predict the RNA ligand poses, and ASP rescoring further improved the ranking of ligand binding poses. Due to the weak correlations (R(2) < 0.3) of existing scoring with experimental binding affinities, we implemented two new RNA-specific scoring functions, iMDLScore1 and iMDLScore2, and obtained better correlations with R(2) = 0.70 and 0.79, respectively. We also proposed a multistep virtual screening approach and demonstrated that rDock:rDock_solv together with iMDLScore2 rescoring obtained the best enrichment on the flexible RNA targets, whereas GOLD:GOLD Fitness combined with rDock_solv rescoring outperformed other methods for rigid RNAs. This study provided practical strategies for RNA modeling and offered new insights into RNA-small molecule interactions for drug discovery.

FINDSITE-metal: integrating evolutionary information and machine learning for structure-based metal-binding site prediction at the proteome level

The rapid accumulation of gene sequences, many of which are hypothetical proteins with unknown function, has stimulated the development of accurate computational tools for protein function prediction with evolution/structure-based approaches showing considerable promise. In this article, we present FINDSITE-metal, a new threading-based method designed specifically to detect metal-binding sites in modeled protein structures. Comprehensive benchmarks using different quality protein structures show that weakly homologous protein models provide sufficient structural information for quite accurate annotation by FINDSITE-metal. Combining structure/evolutionary information with machine learning results in highly accurate metal-binding annotations; for protein models constructed by TASSER, whose average Cα RMSD from the native structure is 8.9 Å, 59.5% (71.9%) of the best of top five predicted metal locations are within 4 Å (8 Å) from a bound metal in the crystal structure. For most of the targets, multiple metal-binding sites are detected with the best predicted binding site at rank 1 and within the top two ranks in 65.6% and 83.1% of the cases, respectively. Furthermore, for iron, copper, zinc, calcium, and magnesium ions, the binding metal can be predicted with high, typically 70% to 90%, accuracy. FINDSITE-metal also provides a set of confidence indexes that help assess the reliability of predictions. Finally, we describe the proteome-wide application of FINDSITE-metal that quantifies the metal-binding complement of the human proteome. FINDSITE-metal is freely available to the academic community at http://cssb.biology.gatech.edu/findsite-metal/.

[Molecular docking: role of intermolecular contacts in formation of complexes of proteins with nucleotides and peptides]

Knowledge of 3D-structure of protein-ligand complex is a major prerequisite for understanding the functioning mechanism of cellular proteins and membrane receptors. This is also of a great help in rational drug design projects. In the present paper we briefly review the molecular docking approaches used to predict possible orientation of a ligand in the protein binding site. The recent trends to improve the accuracy and efficiency of docking algorithms are demonstrated with the results obtained in Laboratory of Biomolecular Modeling. Particular attention is paid to protein-ligand hydrophobic and stacking interactions responsible for molecular recognition of ligand fragments. Such type of interactions are not always adequately represented in scoring criteria of docking applications that leads to mismatch in 3D-structure complexes predictions. That is why further inquiry of methods to account for these interactions is now the area of active research.

The utility of geometrical and chemical restraint information extracted from predicted ligand-binding sites in protein structure refinement

Exhaustive exploration of molecular interactions at the level of complete proteomes requires efficient and reliable computational approaches to protein function inference. Ligand docking and ranking techniques show considerable promise in their ability to quantify the interactions between proteins and small molecules. Despite the advances in the development of docking approaches and scoring functions, the genome-wide application of many ligand docking/screening algorithms is limited by the quality of the binding sites in theoretical receptor models constructed by protein structure prediction. In this study, we describe a new template-based method for the local refinement of ligand-binding regions in protein models using remotely related templates identified by threading. We designed a Support Vector Regression (SVR) model that selects correct binding site geometries in a large ensemble of multiple receptor conformations. The SVR model employs several scoring functions that impose geometrical restraints on the Cα positions, account for the specific chemical environment within a binding site and optimize the interactions with putative ligands. The SVR score is well correlated with the RMSD from the native structure; in 47% (70%) of the cases, the Pearson's correlation coefficient is >0.5 (>0.3). When applied to weakly homologous models, the average heavy atom, local RMSD from the native structure of the top-ranked (best of top five) binding site geometries is 3.1Å (2.9Å) for roughly half of the targets; this represents a 0.1 (0.3)Å average improvement over the original predicted structure. Focusing on the subset of strongly conserved residues, the average heavy atom RMSD is 2.6Å (2.3Å). Furthermore, we estimate the upper bound of template-based binding site refinement using only weakly related proteins to be ∼2.6Å RMSD. This value also corresponds to the plasticity of the ligand-binding regions in distant homologues. The Binding Site Refinement (BSR) approach is available to the scientific community as a web server that can be accessed at http://cssb.biology.gatech.edu/bsr/.

Q-Dock(LHM): Low-resolution refinement for ligand comparative modeling

The success of ligand docking calculations typically depends on the quality of the receptor structure. Given improvements in protein structure prediction approaches, approximate protein models now can be routinely obtained for the majority of gene products in a given proteome. Structure-based virtual screening of large combinatorial libraries of lead candidates against theoretically modeled receptor structures requires fast and reliable docking techniques capable of dealing with structural inaccuracies in protein models. Here, we present Q-Dock(LHM), a method for low-resolution refinement of binding poses provided by FINDSITE(LHM), a ligand homology modeling approach. We compare its performance to that of classical ligand docking approaches in ligand docking against a representative set of experimental (both holo and apo) as well as theoretically modeled receptor structures. Docking benchmarks reveal that unlike all-atom docking, Q-Dock(LHM) exhibits the desired tolerance to the receptor's structure deformation. Our results suggest that the use of an evolution-based approach to ligand homology modeling followed by fast low-resolution refinement is capable of achieving satisfactory performance in ligand-binding pose prediction with promising applicability to proteome-scale applications.

FINDSITE: a combined evolution/structure-based approach to protein function prediction

A key challenge of the post-genomic era is the identification of the function(s) of all the molecules in a given organism. Here, we review the status of sequence and structure-based approaches to protein function inference and ligand screening that can provide functional insights for a significant fraction of the approximately 50% of ORFs of unassigned function in an average proteome. We then describe FINDSITE, a recently developed algorithm for ligand binding site prediction, ligand screening and molecular function prediction, which is based on binding site conservation across evolutionary distant proteins identified by threading. Importantly, FINDSITE gives comparable results when high-resolution experimental structures as well as predicted protein models are used.

Q-Dock: Low-resolution flexible ligand docking with pocket-specific threading restraints

The rapidly growing number of theoretically predicted protein structures requires robust methods that can utilize low-quality receptor structures as targets for ligand docking. Typically, docking accuracy falls off dramatically when apo or modeled receptors are used in docking experiments. Low-resolution ligand docking techniques have been developed to deal with structural inaccuracies in predicted receptor models. In this spirit, we describe the development and optimization of a knowledge-based potential implemented in Q-Dock, a low-resolution flexible ligand docking approach. Self-docking experiments using crystal structures reveals satisfactory accuracy, comparable with all-atom docking. All-atom models reconstructed from Q-Dock's low-resolution models can be further refined by even a simple all-atom energy minimization. In decoy-docking against distorted receptor models with a root-mean-square deviation, RMSD, from native of approximately 3 A, Q-Dock recovers on average 15-20% more specific contacts and 25-35% more binding residues than all-atom methods. To further improve docking accuracy against low-quality protein models, we propose a pocket-specific protein-ligand interaction potential derived from weakly homologous threading holo-templates. The success rate of Q-Dock employing a pocket-specific potential is 6.3 times higher than that previously reported for the Dolores method, another low-resolution docking approach.

Assessment of programs for ligand binding affinity prediction

The prediction of the binding free energy between a ligand and a protein is an important component in the virtual screening and lead optimization of ligands for drug discovery. To determine the quality of current binding free energy estimation programs, we examined FlexX, X-Score, AutoDock, and BLEEP for their performance in binding free energy prediction in various situations including cocrystallized complex structures, cross docking of ligands to their non-cocrystallized receptors, docking of thermally unfolded receptor decoys to their ligands, and complex structures with "randomized" ligand decoys. In no case was there a satisfactory correlation between the experimental and estimated binding free energies over all the datasets tested. Meanwhile, a strong correlation between ligand molecular weight-binding affinity correlation and experimental predicted binding affinity correlation was found. Sometimes the programs also correctly ranked ligands' binding affinities even though native interactions between the ligands and their receptors were essentially lost because of receptor deformation or ligand randomization, and the programs could not decisively discriminate randomized ligand decoys from their native ligands; this suggested that the tested programs miss important components for the accurate capture of specific ligand binding interactions.

Docking of ATP to Ca-ATPase: considering protein domain motions

Most standard molecular docking algorithms take into account only ligand flexibility, while numerous studies demonstrate that receptor flexibility may be also important. While some efficient methods have been proposed to take into account local flexibility of protein side chains, the influence of large-scale domain motions on the docking results still represents a challenge for computational methods. In this work we compared the results of ATP docking to different models of Ca-ATPase: crystallographic apo- and holo-forms of the enzyme as well as "flexible" target models generated via molecular dynamics (MD) simulations in water. MD simulations were performed for two different apo-forms and one holo-form of Ca2+-ATPase and reveal large-scale domain motions of type "closure", which is consistent with experimental structures. Docking to a set of MD-conformers yielded correct solutions with ATP bound in both domains regardless of the starting Ca2+-ATPase structure. Also, special attention was paid to proper ranking of docking solutions and some particular features of different scoring functions and their applicability for the model of "flexible" receptor. Particularly, the results of docking ATP were ranked by a scoring criterion specially designed to estimate ATP-protein interactions. This criterion includes stacking and hydrophobic interactions characteristic of ATP-protein complexes. The performance of this ligand-specific scoring function was considerably better than that of a standard scoring function used in the docking algorithm.

Structure modeling of all identified G protein-coupled receptors in the human genome

G protein–coupled receptors (GPCRs), encoded by about 5% of human genes, comprise the largest family of integral membrane proteins and act as cell surface receptors responsible for the transduction of endogenous signal into a cellular response. Although tertiary structural information is crucial for function annotation and drug design, there are few experimentally determined GPCR structures. To address this issue, we employ the recently developed threading assembly refinement (TASSER) method to generate structure predictions for all 907 putative GPCRs in the human genome. Unlike traditional homology modeling approaches, TASSER modeling does not require solved homologous template structures; moreover, it often refines the structures closer to native. These features are essential for the comprehensive modeling of all human GPCRs when close homologous templates are absent. Based on a benchmarked confidence score, approximately 820 predicted models should have the correct folds. The majority of GPCR models share the characteristic seven-transmembrane helix topology, but 45 ORFs are predicted to have different structures. This is due to GPCR fragments that are predominantly from extracellular or intracellular domains as well as database annotation errors. Our preliminary validation includes the automated modeling of bovine rhodopsin, the only solved GPCR in the Protein Data Bank. With homologous templates excluded, the final model built by TASSER has a global C_α root-mean-squared deviation from native of 4.6 Å, with a root-mean-squared deviation in the transmembrane helix region of 2.1 Å. Models of several representative GPCRs are compared with mutagenesis and affinity labeling data, and consistent agreement is demonstrated. Structure clustering of the predicted models shows that GPCRs with similar structures tend to belong to a similar functional class even when their sequences are diverse. These results demonstrate the usefulness and robustness of the in silico models for GPCR functional analysis. All predicted GPCR models are freely available for noncommercial users on our Web site (http://www.bioinformatics.buffalo.edu/GPCR).

G protein–coupled receptors (GPCRs) are a large superfamily of integral membrane proteins that transduce signals across the cell membrane. Because of the breadth and importance of the physiological roles undertaken by the GPCR family, many of its members are important pharmacological targets. Although the knowledge of a protein's native structure can provide important insight into understanding its function and for the design of new drugs, the experimental determination of the three-dimensional structure of GPCR membrane proteins has proved to be very difficult. This is demonstrated by the fact that there is only one solved GPCR structure (from bovine rhodopsin) deposited in the Protein Data Bank library. In contrast, there are no human GPCR structures in the Protein Data Bank. To address the need for the tertiary structures of human GPCRs, using just sequence information, the authors use a newly developed threading-assembly-refinement method to generate models for all 907 registered GPCRs in the human genome. About 820 GPCRs are anticipated to have correct topology and transmembrane helix arrangement. A subset of the resulting models is validated by comparison with mutagenesis experimental data, and consistent agreement is demonstrated.

A Flexible Docking Procedure for the Exploration of Peptide Binding Selectivity to Known Structures and Homology Models of PDZ Domains

PDZ domains are important scaffolding modules that typically bind to the C-termini of their interaction partners. Several structures of such complexes have been solved, revealing a conserved binding site in the PDZ domain and an extended conformation of the bound peptide. A compendium of information regarding PDZ complexes demonstrates that dissimilar C-terminal peptides bind to the same PDZ domain, and different PDZ domains can bind the same peptides. A detailed understanding of the PDZ-peptide recognition is needed to elucidate this complexity. To this end, we have designed a family of docking protocols for PDZ domains (termed PDZ-DocScheme) that is based on simulated annealing molecular dynamics and rotamer optimization, and is applicable to the docking of long peptides (20-40 rotatable bonds) to both known PDZ structures and to the more complicated problem of homology models of these domains. The resulting protocol reproduces the structures of PDZ complexes with peptides 4-8 amino acids long within 1-2 A from the experimental structure when the docking is performed to the original structure. If the structure of the target PDZ domain is an apo structure or a homology model, the docking protocol yields structures within 3 A in 9 out of 12 test cases. The automated docking procedure PDZ-DocScheme can serve in the generation of a structural context for validation of PDZ domain specificity from mutagenesis and ligand binding data.