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.     

Ensemble docking into flexible active sites. Critical evaluation of FlexE against JNK-3 and beta-secretase

One of the main complicating factors in structure-based drug design is the conformational rearrangement of the receptor upon ligand binding implicating protein flexibility as a crucial component in virtual screening. The FlexE approach allows flexibility through discrete alternative conformations of varying parts of the protein taken from structures having similar backbone traces. Here the performance of FlexE was tested against that of FlexX and FlexX-Pharm, by carrying out virtual screening experiments on two sets of structurally distinct complexes, for the enzymes beta-secretase (BACE), and c-jun N-terminal kinase 3 (JNK-3). A large number of incompatible instances occurred between structural elements of the proteins thus loop movements could not be studied in JNK-3 as well as in BACE. The investigation of the side-chain flexibility revealed that at the most FlexE could achieve the enrichment yielded by FlexX in JNK-3 but not in BACE. Although limited side-chain variations (e.g. different protonation states) can be treated by FlexE, docking into protein ensembles remains a practical tool that decreases the average run time for a ligand.     

SODOCK: swarm optimization for highly flexible protein-ligand docking

Protein-ligand docking can be formulated as a parameter optimization problem associated with an accurate scoring function, which aims to identify the translation, orientation, and conformation of a docked ligand with the lowest energy. The parameter optimization problem for highly flexible ligands with many rotatable bonds is more difficult than that for less flexible ligands using genetic algorithm (GA)-based approaches, due to the large numbers of parameters and high correlations among these parameters. This investigation presents a novel optimization algorithm SODOCK based on particle swarm optimization (PSO) for solving flexible protein-ligand docking problems. To improve efficiency and robustness of PSO, an efficient local search strategy is incorporated into SODOCK. The implementation of SODOCK adopts the environment and energy function of AutoDock 3.05. Computer simulation results reveal that SODOCK is superior to the Lamarckian genetic algorithm (LGA) of AutoDock, in terms of convergence performance, robustness, and obtained energy, especially for highly flexible ligands. The results also reveal that PSO is more suitable than the conventional GA in dealing with flexible docking problems with high correlations among parameters. This investigation also compared SODOCK with four state-of-the-art docking methods, namely GOLD 1.2, DOCK 4.0, FlexX 1.8, and LGA of AutoDock 3.05. SODOCK obtained the smallest RMSD in 19 of 37 cases. The average 2.29 A of the 37 RMSD values of SODOCK was better than those of other docking programs, which were all above 3.0 A.     

Flexible docking of ligands into synthetic receptors using a two-sided incremental construction algorithm

We present a new algorithm for the fast and reliable structure prediction of synthetic receptor-ligand complexes. Our method is based on the protein-ligand docking program FlexX and extends our recently introduced docking technique for synthetic receptors, which has been implemented in the program FlexR. To handle the flexibility of the relevant molecules, we apply a novel docking strategy that uses an adaptive two-sided incremental construction algorithm which incorporates the structural flexibility of both the ligand and synthetic receptor. We follow an adaptive strategy, in which one molecule is expanded by attaching its next fragment in all possible torsion angles, whereas the other (partially assembled) molecule serves as a rigid binding partner. Then the roles of the molecules are exchanged. Geometric filters are used to discard partial conformations that cannot realize a targeted interaction pattern derived in a graph-based precomputation phase. The process is repeated until the entire complex is built up. Our algorithm produces promising results on a test data set comprising 10 complexes of synthetic receptors and ligands. The method generated near-native solutions compared to crystal structures in all but one case. It is able to generate solutions within a couple of minutes and has the potential of being used as a virtual screening tool for searching for suitable guest molecules for a given synthetic receptor in large databases of guests and vice versa.     

AIScore chemically diverse empirical scoring function employing quantum chemical binding energies of hydrogen-bonded complexes

In this work we report on a novel scoring function that is based on the LUDI model and focuses on the prediction of binding affinities. AIScore extends the original FlexX scoring function using a chemically diverse set of hydrogen-bonded interactions derived from extensive quantum chemical ab initio calculations. Furthermore, we introduce an algorithmic extension for the treatment of multifurcated hydrogen bonds (XFurcate). Charged and resonance-assisted hydrogen bond energies and hydrophobic interactions as well as a scaling factor for implicit solvation were fitted to experimental data. To this end, we assembled a set of 101 protein-ligand complexes with known experimental binding affinities. Tightly bound water molecules in the active site were considered to be an integral part of the binding pocket. Compared to the original FlexX scoring function, AIScore significantly improves the prediction of the binding free energies of the complexes in their native crystal structures. In combination with XFurcate, AIScore yields a Pearson correlation coefficient of R P = 0.87 on the training set. In a validation run on the PDBbind test set we achieved an R P value of 0.46 for 799 attractively scored complexes, compared to a value of R P = 0.17 and 739 bound complexes obtained with the FlexX original scoring function. The redocking capability of AIScore, on the other hand, does not fully reach the good performance of the original FlexX scoring function. This finding suggests that AIScore should rather be used for postscoring in combination with the standard FlexX incremental ligand construction scheme.     

Improved FlexX docking using FlexS-determined base fragment placement

We report on a novel hybrid FlexX/FlexS docking approach, whereby the base fragment of the test ligand is chosen by FlexS superposition onto a cocrystallized template ligand and then fed into FlexX for the incremental construction of the final solution. The new approach is tested on the diverse 200 protein-ligand complex dataset that has been previously described for FlexX validation. In total, 62.9% of the complexes can be reproduced at rank 1 by our approach, which compares favorably with 46.9% when using FlexX alone. In addition, we report "cross-docking" experiments in which several receptor structures of complexes with identical proteins have been used for docking all cocrystallized ligands of these complexes. The results show that, in almost all cases, the hybrid approach can acceptably dock a ligand into a foreign receptor structure using a different ligand template, can give solutions where FlexX alone fails, and tends to give solutions that are more accurately positioned.     

Ultrafast de novo docking combining pharmacophores and combinatorics

We report on a successful de novo design approach which relies on the combination of multi-million compound combinatorial docking under receptor-based pharmacophore constraints. Inspired by a rationale by A.R. Leach et al., we document on the unification of two steps into one for ligand assembly. In the original work, fragments known to bind in protein active sites were connected forming novel ligand compounds by means of generic skeleton linkers and following a combinatorial approach. In our approach, the knowledge of fragments binding to the protein has been expressed in terms of a receptor-based pharmacophore definition. The combinatorial linking step is performed in situ during docking, starting from combinatorial libraries. Three sample scenarios growing in size and complexity (combinatorial libraries of 1 million, 1.3 million, and 22.4 million compounds) have been created to illustrate the method. Docking could be accomplished between minutes and several hours depending on the outset; the results were throughout promising. Technically, a module compatibility between FlexX(C) and FlexX-Pharm has been established. The background is explained, and the crucial points from an information scientist's perspective are highlighted.     

An alternative method for the evaluation of docking performance: RSR vs RMSD

A new assessment criterion for docking poses is proposed in which experimental electron density is taken into account when evaluating the ability of docking programs to reproduce experimentally observed binding modes. Three docking programs (Gold, Glide, and Fred) were used to generate poses for a set of 88 protein-ligand complexes for which the crystal structure is known. The new criterion is based on the real space R-factor (RSR), which measures how well a group of atoms-in our case the ligand-fits the experimental electron density by comparing that density to the expected density, calculated from the model (i.e., the predicted ligand pose). The RSR-based measure is compared to the traditional criterion, the root-mean-square distance (RMSD) between the docking pose and the binding configuration in the crystallographic model. The results highlight several shortcomings of the RMSD criterion that do not affect the RSR-based measure. Examples illustrate that the RSR-derived approach allows a more meaningful a posteriori assessment of docking methods and results. Practical implications for docking evaluations and for methodological development work in this field are discussed.     

酪氨酸蛋白磷酸酯酶1B抑制剂的分子对接和三维定量构效关系研究

用一种柔性分子对接方法(FlexX)将12个2-草酰胺苯甲酸类抑制剂和酪氨酸蛋白磷酸酯酶(PTP1B)活性口袋进行分子对接,对接程序预测的抑制剂和酶之间的相互作用能与抑制活性之间有很好的相关性(非线性相关系数R²达0.859),这说明对接结果可以比较准确地预测抑制剂和PTP1B之间的结合模式.然后,将33个同类抑制剂的骨架叠合在分子对接预测的结合构象上,用比较分子力场分析方法(CoMFA)对其进行三维定量活性构效关系研究,得到的CoMFA模型具有很好的统计相关性(交互验证回归系数q²为0.650),并可以准确地预测测试集6个化合物的活性(平均标准偏差为0.177).同时,由CoMFA模型得出的抑制剂改造信息与用FlexX预测的结合模式是一致的,进一步证明我们预测的结合模式是正确的.为研究这类抑制剂和PTP1B的结合模式及对抑制剂进行结构改造提供了信息.     

Protein ligand docking based on empirical method for binding affinity estimation

An empirical protein-ligand binding affinity estimation method, SCORE, was incorporated into a popular docking program, DOCK4. The combined program, ScoreDock, was used to reconstruct the 200 protein-ligand complex structures and found to give good results for the complexes with high binding affinities. A quality assessment method for docking results from ScoreDock was developed based on the whole test set and tested by additionally selected complexes. The method significantly improves the docking accuracy and was shown to be reliable in docking quality assessment. As a docking tool in structural based drug design, ScoreDock can screen out final hits directly based on the predicted negative logarithms of dissociation equilibrium constants of protein-ligand complexes, and can explicitly deal with structure water molecules, as well as metal atoms.     

Knowledge-based interaction fingerprint scoring: a simple method for improving the effectiveness of fast scoring functions

A new method for the postprocessing of docking outputs has been developed, based on encoding putative 3D binding modes (docking solutions) as ligand-protein interactions into simple bit strings, a method analogous to the structural interaction fingerprint. Instead of employing traditional scoring functions, the method uses a series of new, knowledge-based scores derived from the similarity of the bit strings for each docking solution to that of a known reference binding mode. A GOLD docking study was carried out using the Bissantz estrogen receptor antagonist set along with the new scoring method. Superior recovery rates, with up to 2-fold enrichments, were observed when the new knowledge-based scoring was compared to the GOLD fitness score. In addition, top ranking sets of molecules (actives and potential actives or decoys) were structurally diverse with low molecular weights and structural complexities. Principal component analysis and clustering of the fingerprints permits the easy separation of active from inactive binding modes and the visualization of diverse binding modes.     

Alternative to consensus scoring--a new approach toward the qualitative combination of docking algorithms

Since the development of the first docking algorithm in the early 1980s a variety of different docking approaches and tools has been created in order to solve the docking problem. Subsequent studies have shown that the docking performance of most tools strongly depends on the considered target. Thus it is hard to choose the best algorithm in the situation at hand. The docking tools FlexX and AutoDock are among the most popular programs for docking flexible ligands into target proteins. Their analysis, comparison, and combination are the topics of this study. In contrast to standard consensus scoring techniques which integrate different scoring algorithms usually only by their rank, we focus on a more general approach. Our new combined docking workflow-AutoxX-unifies the interaction models of AutoDock and FlexX rather than combining the scores afterward which allows interpretability of the results. The performance of FlexX, AutoDock, and the combined algorithm AutoxX was evaluated on the basis of a test set of 204 structures from the Protein Data Bank (PDB). AutoDock and FlexX show a highly diverse redocking accuracy at the different complexes which assures again the usefulness of taking several docking algorithms into account. With the combined docking the number of complexes reproduced below an rmsd of 2.5 A could be raised by 10. AutoxX had a strong positive effect on several targets. The highest performance increase could be found when redocking 20 protein-ligand complexes of alpha-thrombin, plasmepsin, neuraminidase, and d-xylose isomerase. A decrease was found for gamma-chymotrypsin. The results show that--applied to the right target-AutoxX can improve the docking performance compared to AutoDock and FlexX alone.     

Identifying the binding mode of a molecular scaffold

We describe a method for docking of a scaffold-based series and present its advantages over docking of individual ligands, for determining the binding mode of a molecular scaffold in a binding site. The method has been applied to eight different scaffolds of protein kinase inhibitors (PKI). A single analog of each of these eight scaffolds was previously crystallized with different protein kinases. We have used FlexX to dock a set of molecules that share the same scaffold, rather than docking a single molecule. The main mode of binding is determined by the mode of binding of the largest cluster among the docked molecules that share a scaffold. Clustering is based on our 'nearest single neighbor' method [J. Chem. Inf. Comput. Sci., 43 (2003) 208-217]. Additional criteria are applied in those cases in which more than one significant binding mode is found. Using the proposed method, most of the crystallographic binding modes of these scaffolds were reconstructed. Alternative modes, that have not been detected yet by experiments, could also be identified. The method was applied to predict the binding mode of an additional molecular scaffold that was not yet reported and the predicted binding mode has been found to be very similar to experimental results for a closely related scaffold. We suggest that this approach be used as a virtual screening tool for scaffold-based design processes.     

On the applicability of elastic network normal modes in small-molecule docking

Incorporating backbone flexibility into protein-ligand docking is still a challenging problem. In protein-protein docking, normal mode analysis (NMA) has become increasingly popular as it can be used to describe the collective motions of a biological system, but the question of whether NMA can also be useful in predicting the conformational changes observed upon small-molecule binding has only been addressed in a few case studies. Here, we describe a large-scale study on the applicability of NMA for protein-ligand docking using 433 apo/holo pairs of the Astex data sets. On the basis of sets of the first normal modes from the apo structure, we first generated for each paired holo structure a set of conformations that optimally reproduce its C(α) trace with respect to the underlying normal mode subspace. Using AutoDock, GOLD, and FlexX we then docked the original ligands into these conformations to assess how the docking performance depends on the number of modes used to reproduce the holo structure. The results of our study indicate that, even for such a best-case scenario, the use of normal mode analysis in small-molecule docking is restricted and that a general rule on how many modes to use does not seem to exist or at least is not easy to find.     

Evaluation of docking/scoring approaches: A comparative study based on MMP3 inhibitors

An increasing number of docking/scoring programs are available that use different sampling and scoring algorithms. A reliable scoring function is the crucial element of such approaches. Comparative studies are needed to evaluate their current capabilities. DOCK4 with force field and PMF scoring as well as FlexX were used to evaluate the predictive power of these docking/scoring approaches to identify the correct binding mode of 61 MMP-3 inhibitors in a crystal structure of stromelysin and also to rank them according to their different binding affinities. It was found that DOCK4/PMF scoring performs significantly better than FlexX and DOCK4/FF in both ranking ligands and predicting their binding modes. Most notably, DOCK4/PMF was the only scoring/docking approach that found a significant correlation between binding affinity and predicted score of the docked inhibitors. However, comparing only those cases where the correct binding mode was identified (scoring highest among sampled poses), FlexX showed the best `fine tuning' (lowest rmsd) in predicted binding modes. The results suggest that not so much the sampling procedure but rather the scoring function is the crucial element of a docking program.     

Molecular docking using shape descriptors

Molecular docking explores the binding modes of two interacting molecules. The technique is increasingly popular for studying protein-ligand interactions and for drug design. A fundamental problem problem with molecular docking is that orientation space is very large and grows combinatorially with the number of degrees of freedom of the interacting molecules. Here, we describe and evaluate algorithms that improve the efficiency and accuracy of a shape-based docking method. We use molecular organization and sampling techniques to remove the exponential time dependence on molecular size in docking calculations. The new techniques allow us to study systems that were prohibitively large for the original method. The new algorithms are tested in 10 different protein-ligand systems, including 7 systems where the ligand is itself a protein. In all cases, the new algorithms successfully reproduce the experimentally determined configurations of the ligand in the protein.     

A virtual high throughput screen for high affinity cytochrome P450cam substrates. Implications for in silico prediction of drug metabolism

Structure-based virtual screening techniques require reliable scoring functions to discriminate potential substrates effectively. In this study we compared the performance of GOLD, PMF, DOCK and FlexX scoring functions in FlexX flexible docking to cytochrome P450cam binding site. Crystal structures of protein-substrate complexes were most effectively reproduced by the FlexX/PMF method. On the other hand, the FlexX/GOLD approach provided the best correlation between experimental binding constants and predicted scores. Binding modes selected by the FlexX/PMF approach were rescored by GOLD to obtain a reliable measure of binding energetics. The effectiveness of the FlexX/PMF/GOLD method was demonstrated by the correct classification of 32 out of the 33 experimentally studied compounds and also in a virtual HTS test on a library of 10,000 compounds. Although almost all the available functions were developed to be general, our study on cytochrome P450cam substrates suggests that careful selection or even tailoring the scoring function might increase the prediction power of virtual screens significantly. The FlexX/PMF/GOLD methodology was tested on cytochrome P450 3A4 substrates and inhibitors. This preliminary study revealed that the combined function was able to recognise 334 out of the 345 compounds bound to 3A4.     

Protein–ligand docking with multiple flexible side chains

In this work, we validate and analyze the results of previously published cross docking experiments and classify failed dockings based on the conformational changes observed in the receptors. We show that a majority of failed experiments (i.e. 25 out of 33, involving four different receptors: cAPK, CDK2, Ricin and HIVp) are due to conformational changes in side chains near the active site. For these cases, we identify the side chains to be made flexible during docking calculation by superimposing receptors and analyzing steric overlap between various ligands and receptor side chains. We demonstrate that allowing these side chains to assume rotameric conformations enables the successful cross docking of 19 complexes (ligand all atom RMSD < 2.0 A) using our docking software FLIPDock. The number of side receptor side chains interacting with a ligand can vary according to the ligand's size and shape. Hence, when starting from a complex with a particular ligand one might have to extend the region of potential interacting side chains beyond the ones interacting with the known ligand. We discuss distance-based methods for selecting additional side chains in the neighborhood of the known active site. We show that while using the molecular surface to grow the neighborhood is more efficient than Euclidian-distance selection, the number of side chains selected by these methods often remains too large and additional methods for reducing their count are needed. Despite these difficulties, using geometric constraints obtained from the network of bonded and non-bonded interactions to rank residues and allowing the top ranked side chains to be flexible during docking makes 22 out of 25 complexes successful.     

An improved adaptive genetic algorithm for protein–ligand docking

A new optimization model of molecular docking is proposed, and a fast flexible docking method based on an improved adaptive genetic algorithm is developed in this paper. The algorithm takes some advanced techniques, such as multi-population genetic strategy, entropy-based searching technique with self-adaptation and the quasi-exact penalty. A new iteration scheme in conjunction with above techniques is employed to speed up the optimization process and to ensure very rapid and steady convergence. The docking accuracy and efficiency of the method are evaluated by docking results from GOLD test data set, which contains 134 protein-ligand complexes. In over 66.2% of the complexes, the docked pose was within 2.0 A root-mean-square deviation (RMSD) of the X-ray structure. Docking time is approximately in proportion to the number of the rotatable bonds of ligands.     

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.