New and fast statistical-thermodynamic method for computation of protein-ligand binding entropy substantially improves docking accuracy

We present a novel method to estimate the contributions of translational and rotational entropy to protein-ligand binding affinity. The method is based on estimates of the configurational integral through the sizes of clusters obtained from multiple docking positions. Cluster sizes are defined as the intervals of variation of center of ligand mass and Euler angles in the cluster. Then we suggest a method to consider the entropy of torsional motions. We validate the suggested methods on a set of 135 PDB protein-ligand complexes by comparing the averaged root-mean square deviations (RMSD) of the top-scored ligand docked positions, accounting and not accounting for entropy contributions, relative to the experimentally determined positions. We demonstrate that the method increases docking accuracy by 10-21% when used in conjunction with the AutoDock docking program, thus reducing the percent of incorrectly docked ligands by 1.4-fold to four-fold, so that in some cases the percent of ligands correctly docked to within an RMSD of 2 A is above 90%. We show that the suggested method to account for entropy of relative motions is identical to the method based on the Monte Carlo integration over intervals of variation of center of ligand mass and Euler angles in the cluster.     

Calculations of protein-ligand binding entropy of relative and overall molecular motions

In the context of virtual database screening, calculations of protein-ligand binding entropy of relative and overall molecular motions are challenging, owing to the inherent structural complexity of the ligand binding well in the energy landscape of protein-ligand interactions and computing time limitations. We describe a fast statistical thermodynamic method for estimation the binding entropy to address the challenges. The method is based on the integration of the configurational integral over clusters obtained from multiple docked positions. We apply the method in conjunction with 11 popular scoring functions (AutoDock, ChemScore, DrugScore, D-Score, F-Score, G-Score, LigScore, LUDI, PLP, PMF, X-Score) to evaluate the binding entropy of 100 protein-ligand complexes. The averaged values of binding entropy contribution vary from 6.2 to 9.1 kcal/mol, showing good agreement with literature. We calculate positional sizes and the angular volume of the native ligand wells. The averaged geometric mean of positional sizes in principal directions varies from 0.8 to 1.4 A. The calculated range of angular volumes is 3.3-11.8 rad(2). Then we demonstrate that the averaged six-dimensional volume of the native well is larger than the volume of the most populated non-native well in energy landscapes described by all of 11 scoring functions.     

Development and evaluation of a generic evolutionary method for protein-ligand docking

We have developed a generic evolutionary method with an empirical scoring function for the protein-ligand docking, which is a problem of paramount importance in structure-based drug design. This approach, referred to as the GEMDOCK (Generic Evolutionary Method for molecular DOCKing), combines both continuous and discrete search mechanisms. We tested our approach on seven protein-ligand complexes, and the docked lowest energy structures have root-mean-square derivations ranging from 0.32 to 0.99 A with respect to the corresponding crystal ligand structures. In addition, we evaluated GEMDOCK on crossdocking experiments, in which some complexes with an identical protein used for docking all crystallized ligands of these complexes. GEMDOCK yielded 98% docked structures with RMSD below 2.0 A when the ligands were docked into foreign protein structures. We have reported the validation and analysis of our approach on various search spaces and scoring functions. Experimental results show that our approach is robust, and the empirical scoring function is simple and fast to recognize compounds. We found that if GEMDOCK used the RMSD scoring function, then the prediction accuracy was 100% and the docked structures had RMSD below 0.1 A for each test system. These results suggest that GEMDOCK is a useful tool, and may systematically improve the forms and parameters of a scoring function, which is one of major bottlenecks for molecular recognition.     

An iterative knowledge-based scoring function to predict protein-ligand interactions: II. Validation of the scoring function

We have developed an iterative knowledge-based scoring function (ITScore) to describe protein-ligand interactions. Here, we assess ITScore through extensive tests on native structure identification, binding affinity prediction, and virtual database screening. Specifically, ITScore was first applied to a test set of 100 protein-ligand complexes constructed by Wang et al. (J Med Chem 2003, 46, 2287), and compared with 14 other scoring functions. The results show that ITScore yielded a high success rate of 82% on identifying native-like binding modes under the criterion of rmsd < or = 2 A for each top-ranked ligand conformation. The success rate increased to 98% if the top five conformations were considered for each ligand. In the case of binding affinity prediction, ITScore also obtained a good correlation for this test set (R = 0.65). Next, ITScore was used to predict binding affinities of a second diverse test set of 77 protein-ligand complexes prepared by Muegge and Martin (J Med Chem 1999, 42, 791), and compared with four other widely used knowledge-based scoring functions. ITScore yielded a high correlation of R2 = 0.65 (or R = 0.81) in the affinity prediction. Finally, enrichment tests were performed with ITScore against four target proteins using the compound databases constructed by Jacobsson et al. (J Med Chem 2003, 46, 5781). The results were compared with those of eight other scoring functions. ITScore yielded high enrichments in all four database screening tests. ITScore can be easily combined with the existing docking programs for the use of structure-based drug design.     

Multiple ligand docking by Glide: implications for virtual second-site screening

Performance of Glide was evaluated in a sequential multiple ligand docking paradigm predicting the binding modes of 129 protein-ligand complexes crystallized with clusters of 2-6 cooperative ligands. Three sampling protocols (single precision-SP, extra precision-XP, and SP without scaling ligand atom radii-SP hard) combined with three different scoring functions (GlideScore, Emodel and Glide Energy) were tested. The effects of ligand number, docking order and druglikeness of ligands and closeness of the binding site were investigated. On average 36?% of all structures were reproduced with RMSDs lower than 2??. Correctly docked structures reached 50?% when docking druglike ligands into closed binding sites by the SP hard protocol. Cooperative binding to metabolic and transport proteins can dramatically alter pharmacokinetic parameters of drugs. Analyzing the cytochrome P450 subset the SP hard protocol with Emodel ranking reproduced two-thirds of the structures well. Multiple ligand binding is also exploited by the fragment linking approach in lead discovery settings. The HSP90 subset from real life fragment optimization programs revealed that Glide is able to reproduce the positions of multiple bound fragments if conserved water molecules are considered. These case studies assess the utility of Glide in sequential multiple docking applications.     

Structure-based virtual screening with supervised consensus scoring: evaluation of pose prediction and enrichment factors

Since the evaluation of ligand conformations is a crucial aspect of structure-based virtual screening, scoring functions play significant roles in it. However, it is known that a scoring function does not always work well for all target proteins. When one cannot know which scoring function works best against a target protein a priori, there is no standard scoring method to know it even if 3D structure of a target protein-ligand complex is available. Therefore, development of the method to achieve high enrichments from given scoring functions and 3D structure of protein-ligand complex is a crucial and challenging task. To address this problem, we applied SCS (supervised consensus scoring), which employs a rough linear correlation between the binding free energy and the root-mean-square deviation (rmsd) of a native ligand conformations and incorporates protein-ligand binding process with docked ligand conformations using supervised learning, to virtual screening. We evaluated both the docking poses and enrichments of SCS and five scoring functions (F-Score, G-Score, D-Score, ChemScore, and PMF) for three different target proteins: thymidine kinase (TK), thrombin (thrombin), and peroxisome proliferator-activated receptor gamma (PPARgamma). Our enrichment studies show that SCS is competitive or superior to a best single scoring function at the top ranks of screened database. We found that the enrichments of SCS could be limited by a best scoring function, because SCS is obtained on the basis of the five individual scoring functions. Therefore, it is concluded that SCS works very successfully from our results. Moreover, from docking pose analysis, we revealed the connection between enrichment and average centroid distance of top-scored docking poses. Since SCS requires only one 3D structure of protein-ligand complex, SCS will be useful for identifying new ligands.     

Protein-ligand docking against non-native protein conformers

In the validation of protein-ligand docking protocols, performance is mostly measured against native protein conformers, i.e. each ligand is docked into the protein conformation from the structure that contained that ligand. In real-life applications, however, ligands are docked against non-native conformations of the protein, i.e. the apo structure or a structure of a different protein-ligand complex. Here, we have constructed an extensive test set for assessing docking performance against non-native protein conformations. This new test set is built on the Astex Diverse Set (which we recently constructed for assessing native docking performance) and contains 1112 non-native structures for 65 drug targets. Using the protein-ligand docking program GOLD, the Astex Diverse Set and the new Astex Non-native Set, we established that, whereas docking performance (top-ranked solution within 2 A rmsd of the experimental binding mode) is approximately 80% for native docking, this drops to 61% for non-native docking. A similar drop-off is observed for sampling performance (any solution within 2 A): 91% for native docking vs 72% for non-native docking. No significant differences were observed between docking performance against apo and nonapo structures. We found that, whereas small variations in protein conformation are generally tolerated by our rigid docking protocol, larger protein movements result in a catastrophic drop-off in performance. Some docking performance and nearly all sampling performance can be recovered by considering dockings produced against a small number of non-native structures simultaneously. Docking against non-native structures of complexes containing ligands that are similar to the docked ligand also significantly improves both docking performance and sampling performance.     

Validation studies of the site-directed docking program LibDock

The performance of the site-features docking algorithm LibDock has been evaluated across eight GlaxoSmithKline targets as a follow-up to a broad validation study of docking and scoring software (Warren, G. L.; Andrews, W. C.; Capelli, A.; Clarke, B.; Lalonde, J.; Lambert, M. H.; Lindvall, M.; Nevins, N.; Semus, S. F.; Senger, S.; Tedesco, G.; Walls, I. D.; Woolven, J. M.; Peishoff, C. E.; Head, M. S. J. Med. Chem. 2006, 49, 5912-5931). Docking experiments were performed to assess both the accuracy in reproducing the binding mode of the ligand and the retrieval of active compounds in a virtual screening protocol using both the DJD (Diller, D. J.; Merz, K. M., Jr. Proteins 2001, 43, 113-124) and LigScore2 (Krammer, A. K.; Kirchoff, P. D.; Jiang, X.; Venkatachalam, C. M.; Waldman, M. J. Mol. Graphics Modell. 2005, 23, 395-407) scoring functions. This study was conducted using DJD scoring, and poses were rescored using all available scoring functions in the Accelrys LigandFit module, including LigScore2. For six out of eight targets at least 30% of the ligands were docked within a root-mean-square difference (RMSD) of 2.0 A for the crystallographic poses when the LigScore2 scoring function was used. LibDock retrieved at least 20% of active compounds in the top 10% of screened ligands for four of the eight targets in the virtual screening protocol. In both studies the LigScore2 scoring function enhanced the retrieval of crystallographic poses or active compounds in comparison with the results obtained using the DJD scoring function. The results for LibDock accuracy and ligand retrieval in virtual screening are compared to 10 other docking and scoring programs. These studies demonstrate the utility of the LigScore2 scoring function and that LibDock as a feature directed docking method performs as well as docking programs that use genetic/growing and Monte Carlo driven algorithms.     

Ligand-specific scoring functions: improved ranking of docking solutions

Molecular docking is a powerful computational method that has been widely used in many biomolecular studies to predict geometry of a protein-ligand complex. However, while its conformational search algorithms are usually able to generate correct conformation of a ligand in the binding site, the scoring methods often fail to discriminate it among many false variants. We propose to treat this problem by applying more precise ligand-specific scoring filters to re-rank docking solutions. In this way specific features of interactions between protein and different types of compounds can be implicitly taken into account. New scoring functions were constructed including hydrogen bonds, hydrophobic and hydrophilic complementarity terms. These scoring functions also discriminate ligands by the size of the molecule, the total hydrophobicity, and the number of peptide bonds for peptide ligands. Weighting coefficients of the scoring functions were adjusted using a training set of 60 protein-ligand complexes. The proposed method was then tested on the results of docking obtained for an additional 70 complexes. In both cases the success rate was 5-8% better compared to the standard functions implemented in popular docking software.     

Comparative study of several algorithms for flexible ligand docking

We have performed a comparative assessment of several programs for flexible molecular docking: DOCK 4.0, FlexX 1.8, AutoDock 3.0, GOLD 1.2 and ICM 2.8. This was accomplished using two different studies: docking experiments on a data set of 37 protein-ligand complexes and screening a library containing 10,037 entries against 11 different proteins. The docking accuracy of the methods was judged based on the corresponding rank-one solutions. We have found that the fraction of molecules docked with acceptable accuracy is 0.47, 0.31, 0.35, 0.52 and 0.93 for, respectively, AutoDock, DOCK, FlexX, GOLD and ICM. Thus ICM provided the highest accuracy in ligand docking against these receptors. The results from the other programs are found to be less accurate and of approximately the same quality. A speed comparison demonstrated that FlexX was the fastest and AutoDock was the slowest among the tested docking programs. The database screening was performed using DOCK, FlexX and ICM. ICM was able to identify the original ligands within the top 1% of the total library in 17 cases. The corresponding number for DOCK and FlexX was 7 and 8, respectively. We have estimated that in virtual database screening, 50% of the potentially active compounds will be found among approximately 1.5% of the top scoring solutions found with ICM and among approximately 9% of the top scoring solutions produced by DOCK and FlexX.     

Molecular docking with ligand attached water molecules

A novel approach to incorporate water molecules in protein-ligand docking is proposed. In this method, the water molecules display the same flexibility during the docking simulation as the ligand. The method solvates the ligand with the maximum number of water molecules, and these are then retained or displaced depending on energy contributions during the docking simulation. Instead of being a static part of the receptor, each water molecule is a flexible on/off part of the ligand and is treated with the same flexibility as the ligand itself. To favor exclusion of the water molecules, a constant entropy penalty is added for each included water molecule. The method was evaluated using 12 structurally diverse protein-ligand complexes from the PDB, where several water molecules bridge the ligand and the protein. A considerable improvement in successful docking simulations was found when including flexible water molecules solvating hydrogen bonding groups of the ligand. The method has been implemented in the docking program Molegro Virtual Docker (MVD).     

Development and validation of a modular, extensible docking program: DOCK 5

We report on the development and validation of a new version of DOCK. The algorithm has been rewritten in a modular format, which allows for easy implementation of new scoring functions, sampling methods and analysis tools. We validated the sampling algorithm with a test set of 114 protein-ligand complexes. Using an optimized parameter set, we are able to reproduce the crystal ligand pose to within 2 A of the crystal structure for 79% of the test cases using our rigid ligand docking algorithm with an average run time of 1 min per complex and for 72% of the test cases using our flexible ligand docking algorithm with an average run time of 5 min per complex. Finally, we perform an analysis of the docking failures in the test set and determine that the sampling algorithm is generally sufficient for the binding pose prediction problem for up to 7 rotatable bonds; i.e. 99% of the rigid ligand docking cases and 95% of the flexible ligand docking cases are sampled successfully. We point out that success rates could be improved through more advanced modeling of the receptor prior to docking and through improvement of the force field parameters, particularly for structures containing metal-based cofactors.     

Detailed analysis of grid-based molecular docking: A case study of CDOCKER-A CHARMm-based MD docking algorithm

The influence of various factors on the accuracy of protein-ligand docking is examined. The factors investigated include the role of a grid representation of protein-ligand interactions, the initial ligand conformation and orientation, the sampling rate of the energy hyper-surface, and the final minimization. A representative docking method is used to study these factors, namely, CDOCKER, a molecular dynamics (MD) simulated-annealing-based algorithm. A major emphasis in these studies is to compare the relative performance and accuracy of various grid-based approximations to explicit all-atom force field calculations. In these docking studies, the protein is kept rigid while the ligands are treated as fully flexible and a final minimization step is used to refine the docked poses. A docking success rate of 74% is observed when an explicit all-atom representation of the protein (full force field) is used, while a lower accuracy of 66-76% is observed for grid-based methods. All docking experiments considered a 41-member protein-ligand validation set. A significant improvement in accuracy (76 vs. 66%) for the grid-based docking is achieved if the explicit all-atom force field is used in a final minimization step to refine the docking poses. Statistical analysis shows that even lower-accuracy grid-based energy representations can be effectively used when followed with full force field minimization. The results of these grid-based protocols are statistically indistinguishable from the detailed atomic dockings and provide up to a sixfold reduction in computation time. For the test case examined here, improving the docking accuracy did not necessarily enhance the ability to estimate binding affinities using the docked structures.     

Using buriedness to improve discrimination between actives and inactives in docking

A continuing problem in protein-ligand docking is the correct relative ranking of active molecules versus inactives. Using the ChemScore scoring function as implemented in the GOLD docking software, we have investigated the effect of scaling hydrogen bond, metal-ligand, and lipophilic interactions based on the buriedness of the interaction. Buriedness was measured using the receptor density, the number of protein heavy atoms within 8.0 A. Terms in the scaling functions were optimized using negative data, represented by docked poses of inactive molecules. The objective function was the mean rank of the scores of the active poses in the Astex Diverse Set (Hartshorn et al. J. Med. Chem., 2007, 50, 726) with respect to the docked poses of 99 inactives. The final four-parameter model gave a substantial improvement in the average rank from 18.6 to 12.5. Similar results were obtained for an independent test set. Receptor density scaling is available as an option in the recent GOLD release.     

Rapid protein-ligand costructures using chemical shift perturbations

Structure-based drug discovery requires the iterative determination of protein-ligand costructures in order to improve the binding affinity and selectivity of potential drug candidates. In general, X-ray and NMR structure determination methods are time consuming and are typically the limiting factor in the drug discovery process. The application of molecular docking simulations to filter and evaluate drug candidates has become a common method to improve the throughput and efficiency of structure-based drug design. Unfortunately, molecular docking methods suffer from common problems that include ambiguous ligand conformers or failure to predict the correct docked structure. A rapid approach to determine accurate protein-ligand costructures is described based on NMR chemical shift perturbation (CSP) data routinely obtained using 2D 1H-15N HSQC spectra in high-throughput ligand affinity screens. The CSP data is used to both guide and filter AutoDock calculations using our AutoDockFilter program. This method is demonstrated for 19 distinct protein-ligand complexes where the docked conformers exhibited an average rmsd of 1.17 +/- 0.74 A relative to the original X-ray structures for the protein-ligand complexes.     

The effect of water displacement on binding thermodynamics: concanavalin A

Interactions at the binding interface of biomolecular complexes are often mediated by ordered water molecules. In this work, we considered two concanavalin A-carbohydrate complexes. In the first, a water molecule is buried at the binding interface. In the second, this water molecule is displaced by a modification of the ligand (Clarke, C.; Woods, R. J.; Gluska, J.; Cooper, A.; Nutley, M. A.; Boons, G. J. J. Am. Chem. Soc. 2001, 123, 12238-12247). We computed the contribution of this water molecule to the thermodynamic properties using statistical mechanical formulas for the energy and entropy and molecular dynamics simulations. Other contributions to the binding affinity, including desolvation, entropy of conformational restriction, and interaction between the ligand and protein, were also computed. The thermodynamic consequences of displacement of the ordered water molecule by ligand modification were in qualitative agreement with experimental data. The free energy contribution of the water molecule (-17.2 kcal/mol; -19.2 enthalpic and +2 entropic) was nearly equivalent to the additional protein-ligand interactions in trimannoside 2 (-18.9 kcal/mol). The two structural ions interact more strongly with the water than with the hydroxyl of trimannoside 2, thus favoring trimannoside 1. The contributions from desolvation and conformational entropy are much smaller but significant, compared to the binding free energy difference. The picture that emerges is that the final outcome of water displacement is sensitive to the details of the binding site and cannot be predicted by simple empirical rules.     

Cross-docking of inhibitors into CDK2 structures. 1

Predicting protein/ligand binding affinity is one of the most challenging computational chemistry tasks. Numerous methods have been developed to address this challenge, but they all have limitations. Failure to account for protein flexibility has been a shortcoming of many methods. In this cross-docking study the data set comprised 150 inhibitor complexes of the protein kinase CDK2. Gold and Glide performed well in terms of docking accuracy. The chance of cross-docking a ligand within a 2 A RMSD of its experimental pose was found to be 50%. Relative binding potency was not properly predicted from scoring functions, even though cross-docking of each inhibitor into each protein structure was performed and only scores of correctly docked ligands were considered. An accompanying paper (Voigt, J. H.; Elkin, C.; Madison, V. S. Duca, J. S. J. Chem. Inf. Model. 2008, 48, 669-678) covers cross-docking and docking accuracy from the perspective of using multiple protein structures.     

A novel scoring function for molecular docking

We present a novel scoring function for docking of small molecules to protein binding sites. The scoring function is based on a combination of two main approaches used in the field, the empirical and knowledge-based approaches. To calibrate the scoring function we used an iterative procedure in which a ligand's position and its score were determined self-consistently at each iteration. The scoring function demonstrated superiority in prediction of ligand positions in docking tests against the commonly used Dock, FlexX and Gold docking programs. It also demonstrated good accuracy of binding affinity prediction for the docked ligands.     

A QXP-based multistep docking procedure for accurate prediction of protein-ligand complexes

The two great challenges of the docking process are the prediction of ligand poses in a protein binding site and the scoring of the docked poses. Ligands that are composed of extended chains in their molecular structure display the most difficulties, predominantly because of the torsional flexibility. On the basis of the molecular docking program QXP-Flo+0802, we have developed a procedure particularly for ligands with a high degree of rotational freedom that allows the accurate prediction of the orientation and conformation of ligands in protein binding sites. Starting from an initial full Monte Carlo docking experiment, this was achieved by performing a series of successive multistep docking runs using a local Monte Carlo search with a restricted rotational angle, by which the conformational search space is limited. The method was established by using a highly flexible acetylcholinesterase inhibitor and has been applied to a number of challenging protein-ligand complexes known from the literature.     

Scoring confidence index: statistical evaluation of ligand binding mode predictions

Protein-ligand docking programs can generate a large number of possible binding orientations for each ligand candidate. The challenge is to identify the orientations closest to the native binding mode using a scoring method. Many different scoring functions have been developed for protein-ligand scoring, but their performance on binding mode prediction is often target-dependent. In this study, a statistical approach was employed to provide a confidence measure of scoring performance in finding close to the correct docked ligand orientations. It exploits the fact that the scores provided by an adequately performing scoring function generally improve as the ligand binding modes get closer to the correct native orientation. For such cases, the correlation coefficient of scores versus distances is expected to be highest when the most native-like orientation is used as a reference. This correlation coefficient, called the correlation-based score (CBScore), was used as an indicator of how far the docked pose was from the native orientation. The correlation between the original scores and CBScores as well as the range of CBScores were found to be good measures of scoring performance. They were combined into a single quantity, called the scoring confidence index. High values of the scoring confidence index were indicative of pronounced and relatively smooth binding energy landscapes with easily discernable global minima, resulting in reliable binding mode predictions. Low values of this index reflected rugged energy landscapes making the prediction of the correct binding mode very difficult and often unreliable. The diagnostic ability of the scoring confidence index was tested on a non-redundant set of 50 protein-ligand complexes scored with three commonly employed scoring functions: AffiScore, DrugScore and X-Score. Binding mode predictions were found to be three times more reliable for complexes with scoring confidence indices in the upper half than for cases with values in the lower half of the resulting range of 0–1.6. This new confidence measure of scoring performance is expected to be a valuable tool for virtual screening applications. Electronic supplementary material  The online version of this article (doi:) contains supplementary material, which is available to authorized users.