Docking Performance Research Articles

Docking, scoring, and affinity prediction in CAPRI

We present the fifth evaluation of docking and related scoring methods used in the community-wide experiment on the Critical Assessment of Predicted Interactions (CAPRI). The evaluation examined predictions submitted for a total of 15 targets in eight CAPRI rounds held during the years 2010-2012. The targets represented one the most diverse set tackled by the CAPRI community so far. They included only 10 "classical" docking and scoring problems. In one of the classical targets, the new challenge was to predict the position of water molecules in the protein-protein interface. The remaining five targets represented other new challenges that involved estimating the relative binding affinity and the effect of point mutations on the stability of designed and natural protein-protein complexes. Although the 10 classical CAPRI targets included two difficult multicomponent systems, and a protein-oligosaccharide complex with which CAPRI participants had little experience, this evaluation indicates that the performance of docking and scoring methods has remained quite robust. More remarkably, we find that automatic docking servers exhibit a significantly improved performance, with some servers now performing on par with predictions done by humans. The performance of CAPRI participants in the new challenges, briefly reviewed here, was mediocre overall, but some groups did relatively well and their approaches suggested ways of improving methods for designing binders and for estimating the free energies of protein assemblies, which should impact the field of protein modeling and design as a whole.

Optimization of Compound Ranking for Structure-Based Virtual Ligand Screening Using an Established FRED-Surflex Consensus Approach

The use of multiple target conformers has been applied successfully in virtual screening campaigns; however, a study on how to best combine scores for multiple targets in a hierarchic method that combines rigid and flexible docking is not available. In this study, we used a data set of 59479 compounds to screen multiple conformers of four distinct protein targets to obtain an adapted and optimized combination of an established hierarchic method that employs the programs FRED and Surflex. Our study was extended and verified by application of our protocol to ten different data sets from the directory of useful decoys (DUD). We quantitated overall method performance in ensemble docking and compared several consensus scoring methods to improve the enrichment during virtual ligand screening. We conclude that one of the methods used, which employs a consensus weighted scoring of multiple target conformers, performs consistently better than methods that do not include such consensus scoring. For optimal overall performance in ensemble docking, it is advisable to first calculate a consensus of FRED results and use this consensus as a sub-data set for Surflex screening. Furthermore, we identified an optimal method for each of the chosen targets and propose how to optimize the enrichment for any target.

Evaluation and Optimization of Virtual Screening Workflows with DEKOIS 2.0 – A Public Library of Challenging Docking Benchmark Sets

The application of molecular benchmarking sets helps to assess the actual performance of virtual screening (VS) workflows. To improve the efficiency of structure-based VS approaches, the selection and optimization of various parameters can be guided by benchmarking. With the DEKOIS 2.0 library, we aim to further extend and complement the collection of publicly available decoy sets. Based on BindingDB bioactivity data, we provide 81 new and structurally diverse benchmark sets for a wide variety of different target classes. To ensure a meaningful selection of ligands, we address several issues that can be found in bioactivity data. We have improved our previously introduced DEKOIS methodology with enhanced physicochemical matching, now including the consideration of molecular charges, as well as a more sophisticated elimination of latent actives in the decoy set (LADS). We evaluate the docking performance of Glide, GOLD, and AutoDock Vina with our data sets and highlight existing challenges for VS tools. All DEKOIS 2.0 benchmark sets will be made accessible at http://www.dekois.com.

CSAR Benchmark Exercise 2011–2012: Evaluation of Results from Docking and Relative Ranking of Blinded Congeneric Series

The Community Structure–Activity Resource (CSAR) recently held its first blinded exercise based on data provided by Abbott, Vertex, and colleagues at the University of Michigan, Ann Arbor. A total of 20 research groups submitted results for the benchmark exercise where the goal was to compare different improvements for pose prediction, enrichment, and relative ranking of congeneric series of compounds. The exercise was built around blinded high-quality experimental data from four protein targets: LpxC, Urokinase, Chk1, and Erk2. Pose prediction proved to be the most straightforward task, and most methods were able to successfully reproduce binding poses when the crystal structure employed was co-crystallized with a ligand from the same chemical series. Multiple evaluation metrics were examined, and we found that RMSD and native contact metrics together provide a robust evaluation of the predicted poses. It was notable that most scoring functions underpredicted contacts between the hetero atoms (i.e., N, O, S, etc.) of the protein and ligand. Relative ranking was found to be the most difficult area for the methods, but many of the scoring functions were able to properly identify Urokinase actives from the inactives in the series. Lastly, we found that minimizing the protein and correcting histidine tautomeric states positively trended with low RMSD for pose prediction but minimizing the ligand negatively trended. Pregenerated ligand conformations performed better than those that were generated on the fly. Optimizing docking parameters and pretraining with the native ligand had a positive effect on the docking performance as did using restraints, substructure fitting, and shape fitting. Lastly, for both sampling and ranking scoring functions, the use of the empirical scoring function appeared to trend positively with the RMSD. Here, by combining the results of many methods, we hope to provide a statistically relevant evaluation and elucidate specific shortcomings of docking methodology for the community.

Investigation on the Effect of Key Water Molecules on Docking Performance in CSARdock Exercise

Water molecules are routinely included in molecular docking methods and protocols because of their important role in mediating ligand protein interactions. However, it is still unclear that the inclusion of explicit water molecules improves docking accuracy. To explore the effect of including key water molecules on docking accuracy and performance, we participated in the CSARdock 2011 benchmark exercise. This exercise provides a valuable opportunity for researchers to test their docking programs, methods, and protocols in a blind testing environment. The benchmark exercise and its analysis presented in this paper showed that the performance of current docking programs can be improved by incorporating carefully selected water molecules. Our study showed that water mapping calculations can be used to select key water molecules from experimentally identified water positions for molecular dockings. We have observed that inclusion of all binding site water molecules led to reduced performance and erroneous results. Moreover, an overall improvement in binding pose prediction was achieved when computationally selected water molecules are included during docking simulations. The improvement in the docking performance by including water molecules also depends on protein system, chemical class of ligand, docking method, and scoring function.

Effect of pH and ligand charge state on BACE-1 fragment docking performance

In this work we propose a protocol for estimating the effect of pH on the docking performance to BACE-1, which affords the charge state of the inhibitor as well as the protonation state of all ionisable residues in the protein at a given pH value. To the best of our knowledge, this is the first report of a protocol predicting the BACE-1 ligand docking poses not only at the neutral pH at which most crystallographic structures were obtained, but also at the optimal pH of the enzyme (in the acidic range), at which most of the BACE-1 binding affinity assays are performed. We have applied this protocol to a set of 23 fragment-like BACE-1 ligands that span four orders of magnitude in their binding affinities. The pK a values of the BACE-1 acidic residues deviate substantially from the estimates for model compounds in solution and display a ligand dependent variability, especially in the case of the catalytic Asp dyad residues. This outcome should have a strong bearing on the design of protocols for docking based BACE-1 screening campaigns. Finally, we were able to find an explanation for the poor docking success rate of some fragments based on the availability of anchoring points, a rationale that could help to improve hit rates in BACE-1 screening campaigns.

Use of Experimental Design To Optimize Docking Performance: The Case of LiGenDock, the Docking Module of Ligen, a New De Novo Design Program

On route toward a novel de novo design program, called LiGen, we developed a docking program, LiGenDock, based on pharmacophore models of binding sites, including a non-enumerative docking algorithm. In this paper, we present the functionalities of LiGenDock and its accompanying module LiGenPocket, aimed at the binding site analysis and structure-based pharmacophore definition. We also report the optimization procedure we have carried out to improve the cognate docking and virtual screening performance of LiGenDock. In particular, we applied the design of experiments (DoE) methodology to screen the set of user-adjustable parameters to identify those having the largest influence on the accuracy of the results (which ensure the best performance in pose prediction and in virtual screening approaches) and then to choose their optimal values. The results are also compared with those obtained by two popular docking programs, namely, Glide and AutoDock for pose prediction, and Glide and DOCK6 for Virtual Screening.

Highly SpecIfic and Sensitive Pharmacophore Model for Identifying CXCR4 Antagonists. Comparison with Docking and Shape-Matching Virtual Screening Performance

HIV infection is initiated by fusion of the virus with the target cell through binding of the viral gp120 protein with the CD4 cell surface receptor protein and the CXCR4 or CCR5 coreceptors. There is currently considerable interest in developing novel ligands that can modulate the conformations of these coreceptors and, hence, ultimately block virus-cell fusion. Herein, we present a highly specific and sensitive pharmacophore model for identifying CXCR4 antagonists that could potentially serve as HIV entry inhibitors. Its performance was compared with docking and shape-matching virtual screening approaches using 3OE6 CXCR4 crystal structure and high-affinity ligands as query molecules, respectively. The performance of these methods was compared by virtually screening a library assembled by us, consisting of 228 high affinity known CXCR4 inhibitors from 20 different chemotype families and 4696 similar presumed inactive molecules. The area under the ROC plot (AUC), enrichment factors, and diversity of the resulting virtual hit lists was analyzed. Results show that our pharmacophore model achieves the highest VS performance among all the docking and shape-based scoring functions used. Its high selectivity and sensitivity makes our pharmacophore a very good filter for identifying CXCR4 antagonists.

Docking challenge: protein sampling and molecular docking performance.

Computational tools are essential in the drug design process, especially in order to take advantage of the increasing numbers of solved X-ray and NMR protein-ligand structures. Nowadays, molecular docking methods are routinely used for prediction of protein-ligand interactions and to aid in selecting potent molecules as a part of virtual screening of large databases. The improvements and advances in computational capacity in the past decade have allowed for further developments in molecular docking algorithms to address more complicated aspects such as protein flexibility. The effects of incorporation of active site water molecules and implicit or explicit solvation of the binding site are other relevant issues to be addressed in the docking procedures. Using the right docking algorithm at the right stage of virtual screening is most important. We report a staged study to address the effects of various aspects of protein flexibility and inclusion of active site water molecules on docking effectiveness to retrieve (and to be able to predict) correct ligand poses and to rank docked ligands in relation to their biological activity for CHK1, ERK2, LpxC, and UPA. We generated multiple conformers for the ligand and compared different docking algorithms that use a variety of approaches to protein flexibility, including rigid receptor, soft receptor, flexible side chains, induced fit, and multiple structure algorithms. Docking accuracy varied from 1% to 84%, demonstrating that the choice of method is important.

Feasibility of Using Molecular Docking-Based Virtual Screening for Searching Dual Target Kinase Inhibitors

Multitarget agents have been extensively explored for solving limited efficacies, poor safety, and resistant profiles of an individual target. Theoretical approaches for searching and designing multitarget agents are critically useful. Here, the performance of molecular docking to search dual-target inhibitors for four kinase pairs (CDK2-GSK3B, EGFR-Src, Lck-Src, and Lck-VEGFR2) was assessed. First, the representative structures for each kinase target were chosen by structural clustering of available crystal structures. Next, the performance of molecular docking to distinguish inhibitors from noninhibitors for each individual kinase target was evaluated. The results show that molecular docking-based virtual screening illustrates good capability to find known inhibitors for individual targets, but the prediction accuracy is structurally dependent. Finally, the performance of molecular docking to identify the dual-target kinase inhibitors for four kinase pairs was evaluated. The analyses show that molecular docking successfully filters out most noninhibitors and achieves promising performance for identifying dual-kinase inhibitors for CDK2-GSK3B and Lck-VEGFR2. But a high false-positive rate leads to low enrichment of true dual-target inhibitors in the final list. This study suggests that molecular docking serves as a useful tool in searching inhibitors against dual or even multiple kinase targets, but integration with other virtual screening tools is necessary for achieving better predictions.

How to pick a winning team: approaches towards the selection of computationally derived protein structures for ensemble-based virtual screening

The necessity of treating receptor flexibility in protein-ligand docking has been widely acknowledged and is the subject of extensive research in the field of drug discovery [1]. The use of multiple discrete protein conformations, so-called ensemble docking, has been proven to be a valid concept to mimic target plasticity in docking experiments [2,3]. Using molecular dynamics (MD) the number of different conformations that can be generated is practically unlimited. Not all of these conformations can be included in the pose prediction or virtual screening process for reasons of computational cost. Moreover, some of them will be more suitable for docking purposes than others. The question arises if and how adequate protein conformations can be selected systematically a priori based on quantifiable structural features. For neuraminidase and cyclin-dependent kinase II, snapshots of molecular dynamics simulation trajectories have been clustered and structurally assessed by applying a variety of methods. Extensive cross docking and virtual screening experiments show that relatively large differences in docking performance are caused by only very subtle conformational changes within the protein which cannot be captured by the currently applied characterisation methods. As an alternative, cross docking capability can be used as a reliable indicator towards the selection of suitable conformations for ensemble-based virtual screening. In combination with short minimisations of docked poses in the binding site, virtual screening performance can be further improved and ensembles of MD snapshots can be built which perform as well as the generally superior holo crystal structures.

How to Improve Docking Accuracy of AutoDock4.2: A Case Study Using Different Electrostatic Potentials

Molecular docking, which is the indispensable emphasis in predicting binding conformations and energies of ligands to receptors, constructs the high-throughput virtual screening available. So far, increasingly numerous molecular docking programs have been released, and among them, AutoDock 4.2 is a widely used docking program with exceptional accuracy. It has heretofore been substantiated that the calculation of partial charge is very fundamental for the accurate conformation search and binding energy estimation. However, no systematic comparison of the significances of electrostatic potentials on docking accuracy of AutoDock 4.2 has been determined. In this paper, nine different charge-assigning methods, including AM1-BCC, Del-Re, formal, Gasteiger-Hückel, Gasteiger-Marsili, Hückel, Merck molecular force field (MMFF), and Pullman, as well as the ab initio Hartree-Fock charge, were sufficiently explored for their molecular docking performance by using AutoDock4.2. The results clearly demonstrated that the empirical Gasteiger-Hückel charge is the most applicable in virtual screening for large database; meanwhile, the semiempirical AM1-BCC charge is practicable in lead compound optimization as well as accurate virtual screening for small databases.

ReFlexIn: A Flexible Receptor Protein-Ligand Docking Scheme Evaluated on HIV-1 Protease

For many targets of pharmaceutical importance conformational changes of the receptor protein are relevant during the ligand binding process. A new docking approach, ReFlexIn (Receptor Flexibility by Interpolation), that combines receptor flexibility with the computationally efficient potential grid representation of receptor molecules has been evaluated on the retroviral HIV-1 (Human Immunodeficiency Virus 1) protease system. An approximate inclusion of receptor flexibility is achieved by using interpolation between grid representations of individual receptor conformations. For the retroviral protease the method was tested on an ensemble of protease structures crystallized in the presence of different ligands and on a set of structures obtained from morphing between the unbound and a ligand-bound protease structure. Docking was performed on ligands known to bind to the protease and several non-binders. For the binders the ReFlexIn method yielded in almost all cases ligand placements in similar or closer agreement with experiment than docking to any of the ensemble members without degrading the discrimination with respect to non-binders. The improved docking performance compared to docking to rigid receptors allows for systematic virtual screening applications at very small additional computational cost.

Application of asymmetric statistical potentials to antibody–protein docking

An effective docking algorithm for antibody-protein antigen complex prediction is an important first step toward design of biologics and vaccines. We have recently developed a new class of knowledge-based interaction potentials called Decoys as the Reference State (DARS) and incorporated DARS into the docking program PIPER based on the fast Fourier transform correlation approach. Although PIPER was the best performer in the latest rounds of the CAPRI protein docking experiment, it is much less accurate for docking antibody-protein antigen pairs than other types of complexes, in spite of incorporating sequence-based information on the location of the paratope. Analysis of antibody-protein antigen complexes has revealed an inherent asymmetry within these interfaces. Specifically, phenylalanine, tryptophan and tyrosine residues highly populate the paratope of the antibody but not the epitope of the antigen. Since this asymmetry cannot be adequately modeled using a symmetric pairwise potential, we have removed the usual assumption of symmetry. Interaction statistics were extracted from antibody-protein complexes under the assumption that a particular atom on the antibody is different from the same atom on the antigen protein. The use of the new potential significantly improves the performance of docking for antibody-protein antigen complexes, even without any sequence information on the location of the paratope. We note that the asymmetric potential captures the effects of the multi-body interactions inherent to the complex environment in the antibody-protein antigen interface. The method is implemented in the ClusPro protein docking server, available at http://cluspro.bu.edu.

ALiBERO: evolving a team of complementary pocket conformations rather than a single leader.

Docking and virtual screening (VS) reach maximum potential when the receptor displays the structural changes needed for accurate ligand binding. Unfortunately, these conformational changes are often poorly represented in experimental structures or homology models, debilitating their docking performance. Recently, we have shown that receptors optimized with our LiBERO method (Ligand-guided Backbone Ensemble Receptor Optimization) were able to better discriminate active ligands from inactives in flexible-ligand VS docking experiments. The LiBERO method relies on the use of ligand information for selecting the best performing individual pockets from ensembles derived from normal-mode analysis or Monte Carlo. Here we present ALiBERO, a new computational tool that has expanded the pocket selection from single to multiple, allowing for automatic iteration of the sampling-selection procedure. The selection of pockets is performed by a dual method that uses exhaustive combinatorial search plus individual addition of pockets, selecting only those that maximize the discrimination of known actives compounds from decoys. The resulting optimized pockets showed increased VS performance when later used in much larger unrelated test sets consisting of biologically active and inactive ligands. In this paper we will describe the design and implementation of the algorithm, using as a reference the human estrogen receptor alpha.

High performance transcription factor-DNA docking with GPU computing

BackgroundProtein-DNA docking is a very challenging problem in structural bioinformatics and has important implications in a number of applications, such as structure-based prediction of transcription factor binding sites and rational drug design. Protein-DNA docking is very computational demanding due to the high cost of energy calculation and the statistical nature of conformational sampling algorithms. More importantly, experiments show that the docking quality depends on the coverage of the conformational sampling space. It is therefore desirable to accelerate the computation of the docking algorithm, not only to reduce computing time, but also to improve docking quality.MethodsIn an attempt to accelerate the sampling process and to improve the docking performance, we developed a graphics processing unit (GPU)-based protein-DNA docking algorithm. The algorithm employs a potential-based energy function to describe the binding affinity of a protein-DNA pair, and integrates Monte-Carlo simulation and a simulated annealing method to search through the conformational space. Algorithmic techniques were developed to improve the computation efficiency and scalability on GPU-based high performance computing systems.ResultsThe effectiveness of our approach is tested on a non-redundant set of 75 TF-DNA complexes and a newly developed TF-DNA docking benchmark. We demonstrated that the GPU-based docking algorithm can significantly accelerate the simulation process and thereby improving the chance of finding near-native TF-DNA complex structures. This study also suggests that further improvement in protein-DNA docking research would require efforts from two integral aspects: improvement in computation efficiency and energy function design.ConclusionsWe present a high performance computing approach for improving the prediction accuracy of protein-DNA docking. The GPU-based docking algorithm accelerates the search of the conformational space and thus increases the chance of finding more near-native structures. To the best of our knowledge, this is the first ad hoc effort of applying GPU or GPU clusters to the protein-DNA docking problem.

FRED and HYBRID docking performance on standardized datasets

The docking performance of the FRED and HYBRID programs are evaluated on two standardized datasets from the Docking and Scoring Symposium of the ACS Spring 2011 national meeting. The evaluation includes cognate docking and virtual screening performance. FRED docks 70% of the structures to within 2Å in the cognate docking test. In the virtual screening test, FRED is found to have a mean AUC of 0.75. The HYBRID program uses a modified version of FRED's algorithm that uses both ligand- and structure-based information to dock molecules, which increases its mean AUC to 0.78. HYBRID can also implicitly account for protein flexibility by making use of multiple crystal structures. Using multiple crystal structures improves HYBRID's performance (mean AUC 0.80) with a negligible increase in docking time (~15%).

Docking performance of the glide program as evaluated on the Astex and DUD datasets: a complete set of glide SP results and selected results for a new scoring function integrating WaterMap and glide

Glide SP mode enrichment results for two preparations of the DUD dataset and native ligand docking RMSDs for two preparations of the Astex dataset are presented. Following a best-practices preparation scheme, an average RMSD of 1.140 Å for native ligand docking with Glide SP is computed. Following the same best-practices preparation scheme for the DUD dataset an average area under the ROC curve (AUC) of 0.80 and average early enrichment via the ROC (0.1%) metric of 0.12 were observed. 74 and 56% of the 39 best-practices prepared targets showed AUC over 0.7 and 0.8, respectively. Average AUC was greater than 0.7 for all best-practices protein families demonstrating consistent enrichment performance across a broad range of proteins and ligand chemotypes. In both Astex and DUD datasets, docking performance is significantly improved employing a best-practices preparation scheme over using minimally-prepared structures from the PDB. Enrichment results for WScore, a new scoring function and sampling methodology integrating WaterMap and Glide, are presented for four DUD targets, hivrt, hsp90, cdk2, and fxa. WScore performance in early enrichment is consistently strong and all systems examined show AUC>0.9 and superior early enrichment to DUD best-practices Glide SP results.

The assessment of computationally derived protein ensembles in protein-ligand docking

The inclusion of receptor flexibility in protein-ligand docking experiments has become a major research interest in drug discovery [1,2]. One of the possible methods applied is the use of multiple discrete protein conformations, so called ensemble docking [3,4]. With computational techniques like Molecular Dynamics (MD) a large number of different conformations can be generated, not all of which can or should be included in the docking or virtual screening process [5]. The question arises if and how suitable protein conformations can be selected systematically a priori based on quantifiable conformational features. For neuraminidase and cyclin-dependent kinase II (CDK2), snapshots of MD simulation trajectories have been clustered based on different structural properties using a variety of clustering methods. To establish a possible correlation between docking performance and target conformational attributes the clustered snapshots have been subjected to extensive self- and cross-docking experiments as well as virtual screening using the GOLD docking programme. It is shown that conformationally similar snapshots do not necessarily result in a similar docking or virtual screening performance. The selection of the particular structural property on which to base the clustering appears to be the essential problem.

DecoyFinder: an easy-to-use python GUI application for building target-specific decoy sets

Decoys are molecules that are presumed to be inactive against a target (i.e. will not likely bind to the target) and are used to validate the performance of molecular docking or a virtual screening workflow. The Directory of Useful Decoys database (http://dud.docking.org/) provides a free directory of decoys for use in virtual screening, though it only contains a limited set of decoys for 40 targets.To overcome this limitation, we have developed an application called DecoyFinder that selects, for a given collection of active ligands of a target, a set of decoys from a database of compounds. Decoys are selected if they are similar to active ligands according to five physical descriptors (molecular weight, number of rotational bonds, total hydrogen bond donors, total hydrogen bond acceptors and the octanol-water partition coefficient) without being chemically similar to any of the active ligands used as an input (according to the Tanimoto coefficient between MACCS fingerprints). To the best of our knowledge, DecoyFinder is the first application designed to build target-specific decoy sets. A complete description of the software is included on the application home page. A validation of DecoyFinder on 10 DUD targets is provided as Supplementary Table S1. DecoyFinder is freely available at http://URVnutrigenomica-CTNS.github.com/DecoyFinder.