Dennis M. Krüger

Comparison of Structure- and Ligand-Based Virtual Screening Protocols Considering Hit List Complementarity and Enrichment Factors

Structure‐ and ligand‐based virtual‐screening methods (docking, 2D‐ and 3D‐similarity searching) were analysed for their effectiveness in virtual screening against four different targets: angiotensin‐converting enzyme (ACE), cyclooxygenase 2 (COX‐2), thrombin and human immunodeficiency virus 1 (HIV‐1) protease. The relative performance of the tools was compared by examining their ability to recognise known active compounds from a set of actives and nonactives. Furthermore, we investigated whether the application of different virtual‐screening methods in parallel provides complementary or redundant hit lists. Docking was performed with GOLD, Glide, FlexX and Surflex. The obtained docking poses were rescored by using nine different scoring functions in addition to the scoring functions implemented as objective functions in the docking algorithms. Ligand‐based virtual screening was done with ROCS (3D‐similarity searching), Feature Trees and Scitegic Functional Fingerprints (2D‐similarity searching). The results show that structure‐ and ligand‐based virtual‐screening methods provide comparable enrichments in detecting active compounds. Interestingly, the hit lists that are obtained from different virtual‐screening methods are generally highly complementary. These results suggest that a parallel application of different structure‐ and ligand‐based virtual‐screening methods increases the chance of identifying more (and more diverse) active compounds from a virtual‐screening campaign.

DrugScorePPI webserver: fast and accurate in silico alanine scanning for scoring protein–protein interactions

Protein–protein complexes play key roles in all cellular signal transduction processes. We have developed a fast and accurate computational approach to predict changes in the binding free energy upon alanine mutations in protein–protein interfaces. The approach is based on a knowledge-based scoring function, DrugScorePPI, for which pair potentials were derived from 851 complex structures and adapted against 309 experimental alanine scanning results. Based on this approach, we developed the DrugScorePPI webserver. The input consists of a protein–protein complex structure; the output is a summary table and bar plot of binding free energy differences for wild-type residue-to-Ala mutations. The results of the analysis are mapped on the protein–protein complex structure and visualized using J mol. A single interface can be analyzed within a few minutes. Our approach has been successfully validated by application to an external test set of 22 alanine mutations in the interface of Ras/RalGDS. The DrugScorePPI webserver is primarily intended for identifying hotspot residues in protein–protein interfaces, which provides valuable information for guiding biological experiments and in the development of protein–protein interaction modulators. The DrugScorePPI Webserver, accessible at http://cpclab.uni-duesseldorf.de/dsppi, is free and open to all users with no login requirement.

NMSim Web Server: integrated approach for normal mode-based geometric simulations of biologically relevant conformational transitions in proteins

The NMSim web server implements a three-step approach for multiscale modeling of protein conformational changes. First, the protein structure is coarse-grained using the FIRST software. Second, a rigid cluster normal-mode analysis provides low-frequency normal modes. Third, these modes are used to extend the recently introduced idea of constrained geometric simulations by biasing backbone motions of the protein, whereas side chain motions are biased toward favorable rotamer states (NMSim). The generated structures are iteratively corrected regarding steric clashes and stereochemical constraint violations. The approach allows performing three simulation types: unbiased exploration of conformational space; pathway generation by a targeted simulation; and radius of gyration-guided simulation. On a data set of proteins with experimentally observed conformational changes, the NMSim approach has been shown to be a computationally efficient alternative to molecular dynamics simulations for conformational sampling of proteins. The generated conformations and pathways of conformational transitions can serve as input to docking approaches or more sophisticated sampling techniques. The web server output is a trajectory of generated conformations, Jmol representations of the coarse-graining and a subset of the trajectory and data plots of structural analyses. The NMSim webserver, accessible at http://www.nmsim.de, is free and open to all users with no login requirement.

CNA web server: rigidity theory-based thermal unfolding simulations of proteins for linking structure, (thermo-)stability, and function

The Constraint Network Analysis (CNA) web server provides a user-friendly interface to the CNA approach developed in our laboratory for linking results from rigidity analyses to biologically relevant characteristics of a biomolecular structure. The CNA web server provides a refined modeling of thermal unfolding simulations that considers the temperature dependence of hydrophobic tethers and computes a set of global and local indices for quantifying biomacromolecular stability. From the global indices, phase transition points are identified where the structure switches from a rigid to a floppy state; these phase transition points can be related to a protein’s (thermo-)stability. Structural weak spots (unfolding nuclei) are automatically identified, too; this knowledge can be exploited in data-driven protein engineering. The local indices are useful in linking flexibility and function and to understand the impact of ligand binding on protein flexibility. The CNA web server robustly handles small-molecule ligands in general. To overcome issues of sensitivity with respect to the input structure, the CNA web server allows performing two ensemble-based variants of thermal unfolding simulations. The web server output is provided as raw data, plots and/or Jmol representations. The CNA web server, accessible at http://cpclab.uni-duesseldorf.de/cna or http://www.cnanalysis.de, is free and open to all users with no login requirement.

Elastic Potential Grids: Accurate and Efficient Representation of Intermolecular Interactions for Fully Flexible Docking

Protein–ligand docking is the major workhorse in computeraided structure-based lead finding and optimization. Predicted protein–ligand complex configurations are used for studying protein–ligand interactions, estimating binding affinities, and as a final filter step in virtual screening. Early methods on protein–ligand docking treated either both proteins and ligands as rigid molecules or allowed for conformational flexibility of only the ligand, following a “rigid receptor hypothesis”. However, pronounced plasticity upon ligand binding has been observed for several pharmacologically important proteins, such as HIV-1 protease, aldose reductase, FK506 binding protein, renin, and dihydrofolate reductase (DHFR). Protein plasticity comprises a range of possible movements, from single side chains to drastic structural rearrangements as seen in calmodulin. Not surprisingly, if docking is performed with the assumption of a rigid active site in those cases, a dramatic decrease in docking accuracy is observed: Whereas a docking success rate of 76% was reported for docking a ligand back to the protein structure derived from the ligand’s co-crystal structure (“re-docking”), this rate dropped to only 49% if the ligands were docked against protein structures derived from other ligands’ co-crystal structures (“cross docking”). Similar drop-offs have also been reported by others. Furthermore, the drop in docking accuracy was found to be mirrored by the degree to which the protein moves upon ligand binding 32] so that docking to an empty form (“apo docking”) usually shows the largest deterioration. This clearly highlights the importance of developing strategies for taking protein plasticity into account in addition to the conformational flexibility of the ligand (henceforth referred to as “fully flexible docking”) to prevent mis-dockings of ligands to flexible proteins. At present, three major routes to include protein plasticity during docking can be identified. The classification correlates with various types of protein movements observed upon ligand binding. First, plasticity is considered implicitly following a soft-docking strategy with attenuated repulsive forces between protein and ligand. While this is simple to implement and does not compromise docking efficiency, the range of possible movements that can be covered is rather limited. Second, only side chain conformational changes in the binding pocket are modeled. These approaches assume that the protein has a rigid backbone structure, thus neglecting critical backbone shifts responsible for mis-docking of ligands. Third, large-scale conformational changes including backbone motions are taken into account. There are several types of approaches in this category: perform parallel docking into multiple protein conformations; structurally combine multiple conformations; model protein motions in reduced coordinates; apply molecular dynamics or Monte Carlo based sampling to either generate protein–ligand configurations 50] or optimize pre-computed configurations. Docking accuracy and computational efficiency determine the scope and quality of a docking approach. As for the first, fully flexible docking should ultimately become as accurate as “re-docking” pursued with a “rigid receptor hypothesis”. Preserving computational efficiency is equally important, given the short timeframe usually available for a docking run. In particular, evaluating the interaction energy between protein and ligand is expensive. A widely used approach to increase the calculation speed is based on potential fields that are pre-calculated just once in the binding pocket region of the protein, by scanning interactions between the protein and ligand atom probes. The potential field values are stored at the intersections of a regular 3D grid, providing a lookup table. The approach is applicable to all distance-dependent pairwise interactions, such as electrostatic and van der Waals interactions and interactions described by statistical pair potentials. In subsequent docking runs, interaction energies between protein and ligand are then determined in constant time from the lookup table by means of interpolation. This provides a significant rate increase relative to individually evaluating the pair interactions. However, this regular 3D grid-based approach is incompatible with fully flexible docking, because the lookup table values would need to be recalculated for every new protein conformation considered. In the present study, we therefore developed an accurate representation of intermolecular interactions that makes use of the high efficiency in evaluating protein–ligand interaction energies from lookup tables even in the case of a moving protein. The new lookup table function for potential fields that we introduce is based on irregular, deformable 3D grids (Figure 1). The underlying idea is to adapt a 3D grid with pre-calculated potential field values, which were derived from an initial protein conformation, to another conformation by moving intersection points in space, but keeping the potential field values constant. As in the case of a regular 3D grid, interaction energies between ligand and protein are then determined from this lookup table. In contrast to the established approach, however, new protein conformations can now be sampled [a] S. Kazemi, D. M. Kr ger, Prof. Dr. H. Gohlke Institut f r Pharmazeutische und Medizinische Chemie Heinrich-Heine-Universit t Universit tsstr. 1, 40225 D sseldorf (Germany) Fax: (+49)211-81-13847 E-mail : gohlke@uni-duesseldorf.de [b] Dr. F. Sirockin Novartis Pharma AG, 4002 Basel (Switzerland) Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/cmdc.200900146.

Target Flexibility in RNA-Ligand Docking Modeled by Elastic Potential Grids.

The highly flexible nature of RNA provides a formidable challenge for structure-based drug design approaches that target RNA. We introduce an approach for modeling target conformational changes in RNA-ligand docking based on potential grids that are represented as elastic bodies using Naviers equation. This representation provides an accurate and efficient description of RNA-ligand interactions even in the case of a moving RNA structure. When applied to a data set of 17 RNA-ligand complexes, filtered out of the largest validation data set used for RNA-ligand docking so far, the approach is twice as successful as docking into an apo structure and still half as successful as redocking to the holo structure. The approach allows considering RNA movements of up to 6 Å rmsd and is based on a uniform and robust parametrization of the properties of the elastic potential grids, so that the approach is applicable to different RNA-ligand complex classes.

DrugScorePPI Knowledge-Based Potentials Used as Scoring and Objective Function in Protein-Protein Docking

The distance-dependent knowledge-based DrugScorePPI potentials, previously developed for in silico alanine scanning and hot spot prediction on given structures of protein-protein complexes, are evaluated as a scoring and objective function for the structure prediction of protein-protein complexes. When applied for ranking “unbound perturbation” (“unbound docking”) decoys generated by Baker and coworkers a 4-fold (1.5-fold) enrichment of acceptable docking solutions in the top ranks compared to a random selection is found. When applied as an objective function in FRODOCK for bound protein-protein docking on 97 complexes of the ZDOCK benchmark 3.0, DrugScorePPI/FRODOCK finds up to 10% (15%) more high accuracy solutions in the top 1 (top 10) predictions than the original FRODOCK implementation. When used as an objective function for global unbound protein-protein docking, fair docking success rates are obtained, which improve by ∼2-fold to 18% (58%) for an at least acceptable solution in the top 10 (top 100) predictions when performing knowledge-driven unbound docking. This suggests that DrugScorePPI balances well several different types of interactions important for protein-protein recognition. The results are discussed in view of the influence of crystal packing and the type of protein-protein complex docked. Finally, a simple criterion is provided with which to estimate a priori if unbound docking with DrugScorePPI/FRODOCK will be successful.

Predicting protein-protein interactions with DrugScorePPI: fully-flexible docking, scoring, and in silico alanine-scanning

Protein-protein complexes play key roles in all cellular signal transduction processes. Here, we present a fast and accurate computational approach to predict protein-protein interactions. The approach is based on DrugScorePPI, a knowledge-based scoring function for which pair potentials were derived from 851 complex structures and adapted against 309 experimental alanine scanning results. We developed the DrugScorePPI webserver [1], accessible at http://cpclab.uni-duesseldorf.de/dsppi, that is intended for identifying hotspot residues in protein-protein interfaces. For this, it allows performing computational alanine scanning of a protein-protein interface within a few minutes. Our approach has been successfully validated by application to an external test set of 22 alanine mutations in the interface of Ras/RalGDS and outperformed the widely used CC/PBSA, FoldX, and Robetta methods [1]. Next, DrugScorePPI was teamed with FRODOCK [2], a fast FFT-based protein-protein docking tool, in order to predict 3D structures of protein-protein complexes. When applied to datasets of 54 bound-bound (I) and 54 unbound-unbound (II) test cases, convincing results were obtained (docking success rate for complexes with rmsd < 10 A: I: ~80%; II: ~50%). Thus, we set out to evaluate whether our approach of deformable potential grids [3], previously developed for protein-ligand docking, also provides an accurate and efficient means for representing intermolecular interactions in fully-flexible protein-protein docking. The underlying idea is to adapt a 3D grid of potential field values, pre-calculated from an initial protein conformation by DrugScorePPI, to another conformation by moving grid intersection points in space, but keeping the potential field values constant. Protein movements are thereby translated into grid intersection displacements by coupling protein atoms to nearby grid intersection points by means of harmonic springs and modelling the irregular, deformable 3D grid as a homogeneous linear elastic body applying elasticity theory. Thus, new protein conformations can be sampled during a docking run without the need to re-calculate potential field values.

Towards targeting protein-protein interfaces with small molecules

A promising way to interfere with biological processes is through the control of protein-protein interactions by means of small molecules that modulate the formation of protein-protein complexes. Although the feasibility of this approach has been demonstrated in principle by recent results, many of the small-molecule modulators known to date have not been found by rational design approaches. In large part this is due to the challenges that one faces in dealing with protein binding epitopes compared to, e.g., enzyme binding pockets. Recent advances in the understanding of the energetics and dynamics of protein binding interfaces[1] and methodological developments in the field of structure-based drug design methods may open up a way to apply rational design approaches also for finding protein-protein interaction modulators.2 Here, we first show in a retrospective analysis of the well-investigated interleukin-2 system how I) potential binding sites in an interface can be identified from an unbound protein structure, II) the interface can be dissected in terms of energetic contributions of single residues, and III) one can make use of this knowledge for guiding the development of small-molecule modulators. When applied to a leukaemia-associated fusion protein in a prospective manner, the predictive character of the methodology is demonstrated [2]. Another challenge arises from the fact that protein-protein interfaces are flexible. In the second part, we thus demonstrate a novel approach for including protein flexibility into protein-ligand docking[3]. This approach is based on elastic potential grids, which provide an accurate and efficient representation of intermolecular interactions in fully-flexible docking.