A fast and efficient method to generate biologically relevant conformations

Summary Mutual binding between a ligand of low molecular weight and its macromolecular receptor demands structural complementarity of both species at the recognition site. To predict binding properties of new molecules before synthesis, information about possible conformations of drug molecules at the active site is required, especially if the 3D structure of the receptor is not known. The statistical analysis of small-molecule crystal data allows one to elucidate conformational preferences of molecular fragments and accordingly to compile libraries of putative ligand conformations. A comparison of geometries adopted by corresponding fragments in ligands bound to proteins shows similar distributions in conformation space. We have developed an automatic procedure that generates different conformers of a given ligand. The entire molecule is decomposed into its individual ring and open-chain torsional fragments, each used in a variety of favorable conformations. The latter ones are produced according to the library information about conformational preferences. During this building process, an extensive energy ranking is applied. Conformers ranked as energetically favorable are subjected to an optimization in torsion angle space. During minimization, unfavorable van der Waals interactions are removed while keeping the open-chain torsion angles as close as possible to the experimentally most frequently observed values. In order to assess how well the generated conformers map conformation space, a comparison with experimental data has been performed. This comparison gives some confidence in the efficiency and completeness of this approach. For some ligands that had been structurally characterized by protein crystallography, the program was used to generate sets of some 10 to 100 conformers. Among these, geometries are found that fall convincingly close to the conformations actually adopted by these ligands at the binding site.     

Torsion angle preference and energetics of small-molecule ligands bound to proteins

Small organic molecules can assume conformations in the protein-bound state that are significantly different from those in solution. We have analyzed the conformations of 21 common torsion motifs of small molecules extracted from crystal structures of protein-ligand complexes and compared them with their torsion potentials calculated by an ab initio DFT method. We find a good correlation between the potential energy of the torsion motifs and their conformational distribution in the protein-bound state: The most probable conformations of the torsion motifs agree well with the calculated global energy minima, and the lowest torsion-energy state becomes increasingly dominant as the torsion barrier height increases. The torsion motifs can be divided into 3 groups based on torsion barrier heights: high (>4 kcal/mol), medium (2-4 kcal/mol), and low (<2 kcal/mol). The calculated torsion energy profiles are predictive for the most preferred bound conformation for the high and medium barrier groups, the latter group common in druglike molecules. In the high-barrier group of druglike ligands, >95% of conformational torsions occur in the energy region <4 kcal/mol. The conformations of the torsion motifs in the protein-bound state can be modeled by a Boltzmann distribution with a temperature factor much higher than room temperature. This high-temperature factor, derived by fitting the theoretical model to the experimentally observed conformation occurrence of torsions, can be interpreted as the perturbation that proteins inflict on the conformation of the bound ligand. Using this model, it is calculated that the average strain energy of a torsion motif in ligands bound to proteins is approximately 0.6 kcal/mol, a result which can be related to the lower binding efficiency of larger ligands with more rotatable bonds. The above results indicate that torsion potentials play an important role in dictating ligand conformations in both the free and the bound states.     

Protein:Ligand binding free energies: A stringent test for computational protein design

A computational protein design method is extended to allow Monte Carlo simulations where two ligands are titrated into a protein binding pocket, yielding binding free energy differences. These provide a stringent test of the physical model, including the energy surface and sidechain rotamer definition. As a test, we consider tyrosyl‐tRNA synthetase (TyrRS), which has been extensively redesigned experimentally. We consider its specificity for its substrate l ‐tyrosine (l ‐Tyr), compared to the analogs d ‐Tyr, p‐acetyl‐, and p‐azido‐phenylalanine (ac‐Phe, az‐Phe). We simulate l ‐ and d ‐Tyr binding to TyrRS and six mutants, and compare the structures and binding free energies to a more rigorous “MD/GBSA” procedure: molecular dynamics with explicit solvent for structures and a Generalized Born + Surface Area model for binding free energies. Next, we consider l ‐Tyr, ac‐ and az‐Phe binding to six other TyrRS variants. The titration results are sensitive to the precise rotamer definition, which involves a short energy minimization for each sidechain pair to help relax bad contacts induced by the discrete rotamer set. However, when designed mutant structures are rescored with a standard GBSA energy model, results agree well with the more rigorous MD/GBSA. As a third test, we redesign three amino acid positions in the substrate coordination sphere, with either l ‐Tyr or d ‐Tyr as the ligand. For two, we obtain good agreement with experiment, recovering the wildtype residue when l ‐Tyr is the ligand and a d ‐Tyr specific mutant when d ‐Tyr is the ligand. For the third, we recover His with either ligand, instead of wildtype Gln. © 2015 Wiley Periodicals, Inc.     

Receptor‐Bound Conformation of Cilengitide Better Represented by Its Solution‐State Structure than the Solid‐State Structure

The X‐ray crystal and NMR spectroscopic structures of the peptide drug candidate Cilengitide (cyclo(RGDf(NMe)Val)) in various solvents are obtained and compared in addition to the integrin receptor bound conformation. The NMR‐based solution structures exhibit conformations closely resembling the X‐ray structure of Cilengitide bound to the head group of integrin αvβ3. In contrast, the structure of pure Cilengitide recrystallized from methanol reveals a different conformation controlled by the lattice forces of the crystal packing. Molecular modeling studies of the various ligand structures docked to the αvβ3 integrin revealed that utilization of the solid‐state conformation of Cilengitide leads—unlike the solution‐based structures—to a mismatch of the ligand–receptor interactions compared with the experimentally determined structure of the protein–ligand complex. Such discrepancies between solution and crystal conformations of ligands can be misleading during the structure‐based lead optimization process and should thus be taken carefully into account in ligand orientated drug design.     

Placement of medium-sized molecular fragments into active sites of proteins

Summary We present an algorithm for placing molecular fragments into the active site of a receptor. A molecular fragment is defined as a connected part of a molecule containing only complete ring systems. The algorithm is part of a docking tool, called FlexX, which is currently under development at GMD. The overall goal is to provide means of automatically computing low-energy conformations of the ligand within the active site, with an accuracy approaching the limitations of experimental methods for resolving molecular structures and within a run time that allows for docking large sets of ligands. The methods by which we plan to achieve this goal are the explicit exploitation of molecular flexibility of the ligand and the incorporation of physicochemical properties of the molecules. The algorithm for fragment placement, which is the topic of this paper, is based on pattern recognition techniques and is able to predict a small set of possible positions of a molecular fragment with low flexibility within seconds on a workstation. In most cases, a placement with rms deviation below 1.0 Å with respect to the X-ray structure is found among the 10 highest ranking solutions, assuming that the receptor is given in the bound conformation.     

The MPSim-Dock hierarchical docking algorithm: application to the eight trypsin inhibitor cocrystals

To help improve the accuracy of protein-ligand docking as a useful tool for drug discovery, we developed MPSim-Dock, which ensures a comprehensive sampling of diverse families of ligand conformations in the binding region followed by an enrichment of the good energy scoring families so that the energy scores of the sampled conformations can be reliably used to select the best conformation of the ligand. This combines elements of DOCK4.0 with molecular dynamics (MD) methods available in the software, MPSim. We test here the efficacy of MPSim-Dock to predict the 64 protein-ligand combinations formed by starting with eight trypsin cocrystals, and crossdocking the other seven ligands to each protein conformation. We consider this as a model for how well the method would work for one given target protein structure. Using as a criterion that the structures within 2 kcal/mol of the top scoring include a conformation within a coordinate root mean square (CRMS) of 1 A of the crystal structure, we find that 100% of the 64 cases are predicted correctly. This indicates that MPSim-Dock can be used reliably to identify strongly binding ligands, making it useful for virtual ligand screening.     

Docking flexible molecules: A case study of three proteins

A genetic algorithm (GA) conformation search method is used to dock a series of flexible molecules into one of three proteins. The proteins examined are thermolysin (tmn), carboxypeptidase A (cpa), and dihydrofolate reductase (dfr). In the latter two proteins, the crystal ligand was redocked. For thermolysin, we docked eight ligands into a protein conformation derived from a single crystal structure. The bound conformations of the other ligands in tmn are known. In the cpa and dfr cases, and in seven of the eight tmn ligands, the GA docking method found conformations within 1.6 Å root mean square (rms) of the relaxed crystal conformation. © 1995 John Wiley & Sons, Inc.     

A molecular mechanics/grid method for evaluation of ligand–receptor interactions

We present a computational method for prediction of the conformation of a ligand when bound to a macromolecular receptor. The method is intended for use in systems in which the approximate location of the binding site is known and no large-scale rearrangements of the receptor are expected upon formation of the complex. The ligand is initially placed in the vicinity of the binding site and the atomic motions of the ligand and binding site are explicitly simulated, with solvent represented by an implicit solvation model and using a grid representation for the bulk of the receptor protein. These two approximations make the method computationally efficient and yet maintain accuracy close to that of an all-atom calculation. For the benzamidine/trypsin system, we ran 100 independent simulations, in many of which the ligand settled into the low-energy conformation observed in the crystal structure of the complex. The energy of these conformations was lower than and well-separated from that of others sampled. Extensions of this method are also discussed. © 1995 by John Wiley & Sons, Inc.     

Computational modeling of a binding conformation of the intermediate L-histidinal to histidinol dehydrogenase

Histidinol dehydrogenase (HDH) is one of the enzymes involved in the L-histidine biosynthesis pathway. HDH is a dimer that contains one Zn2+ ion in each identical subunit. In this study, we predicted a possible binding conformation of the intermediate L-histidinal, which is experimentally not known, using a computational modeling method and three potent HDH inhibitors whose structures are similar to that of L-histidinal. At first, a set of the most probable active conformations of the potent inhibitors was determined using two different pharmacophore mapping techniques, the active analogue approach and the distance comparison method. From the most probable active conformations of the three potent inhibitors, the common parts of the L-histidinal structure were extracted and refined by energy minimization to obtain the binding conformation of L-histidinal. This predicted conformation of L-histidinal agrees with an experimentally determined conformation of L-histidine in a single crystal, suggesting that it is an experimentally acceptable conformation. The capability in this conformation to coordinate a Zn2+ ion was examined by comparing the spatial relative geometry of its functional groups with those of ligands that coordinate with a Zn2+ ion in Zn proteins of the Protein Data Bank. This comparison supported our predicted conformation.     

A normal mode-based geometric simulation approach for exploring biologically relevant conformational transitions in proteins

A three-step approach for multiscale modeling of protein conformational changes is presented that incorporates information about preferred directions of protein motions into a geometric simulation algorithm. The first two steps are based on a rigid cluster normal-mode analysis (RCNMA). Low-frequency normal modes are used in the third step (NMSim) to extend the recently introduced idea of constrained geometric simulations of diffusive motions in proteins by biasing backbone motions of the protein, whereas side-chain motions are biased toward favorable rotamer states. 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. When applied to a data set of proteins with experimentally observed conformational changes, conformational variabilities are reproduced very well for 4 out of 5 proteins that show domain motions, with correlation coefficients r > 0.70 and as high as r = 0.92 in the case of adenylate kinase. In 7 out of 8 cases, NMSim simulations starting from unbound structures are able to sample conformations that are similar (root-mean-square deviation = 1.0-3.1 ?) to ligand bound conformations. An NMSim generated pathway of conformational change of adenylate kinase correctly describes the sequence of domain closing. The NMSim approach is 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 as starting points for more sophisticated sampling techniques.     

Validated ligand mapping of ACE active site

Crystal structures of angiotensin-converting enzyme (ACE) complexed with three inhibitors (lisinopril, captopril, enalapril) provided experimental data for testing the validity of a prior active site model predicting the bound conformation of the inhibitors. The ACE active site model - predicted over 18 years ago using a series of potent ACE inhibitors of diverse chemical structure - was recreated using published data and commercial software. Comparison between the predicted structures of the three inhibitors bound to the active site of ACE and those determined experimentally yielded root mean square deviation (RMSD) values of 0.43-0.81 A, among the distances defining the active site map. The bound conformations of the chemically relevant atoms were accurately deduced from the geometry of ligands, applying the assumption that the geometry of the active site groups responsible for binding and catalysis of amide hydrolysis was constrained. The mapping of bound inhibitors at the ACE active site was validated for known experimental compounds, so that the constrained conformational search methodology may be applied with confidence when no experimentally determined structure of the enzyme yet exists, but potent, diverse inhibitors are available.     

Docking small ligands in flexible binding sites

A novel procedure for docking ligands in a flexible binding site is presented. It relies on conjugate gradient minimization, during which nonbonded interactions are gradually switched on. Short Monte Carlo minimization runs are performed on the most promising candidates. Solvation is implicitly taken into account in the evaluation of structures with a continuum model. It is shown that the method is very accurate and can model induced fit in the ligand and the binding site. The docking procedure has been successfully applied to three systems. The first two are the binding of progesterone and 5β-androstane-3,17-dione to the antigen binding fragment of a steroid binding antibody. A comparison of the crystal structures of the free and the two complexed forms reveals that any attempt to model binding must take protein rearrangements into account. Furthermore, the two ligands bind in two different orientations, posing an additional challenge. The third test case is the docking of N^α-(2-naphthyl-sulfonyl-glycyl)-D -para-amidino-phenyl-alanyl-piperidine (NAPAP) to human α-thrombin. In contrast to steroids, NAPAP is a very flexible ligand, and no information of its conformation in the binding site is used. All docking calculations are started from X-ray conformations of proteins with the uncomplexed binding site. For all three systems the best minima in terms of free energy have a root mean square deviation from the X-ray structure smaller than 1.5 Å for the ligand atoms. © 1998 John Wiley & Sons, Inc. J Comput Chem 19: 21–37, 1998     

Conformational selection of protein kinase A revealed by flexible-ligand flexible-protein docking

Protein kinases have high structural plasticity: their structure can change significantly, depending on what ligands are bound to them. Rigid-protein docking methods are not capable of describing such effects. Here, we present a new flexible-ligand flexible-protein docking model in which the protein can adopt conformations between two extremes observed experimentally. The model utilized a molecular dynamics-based simulated annealing cycling protocol and a distance-dependent dielectric model to perform docking. By testing this model on docking four diverse ligands to protein kinase A, we found that the ligands were able to dock successfully to the protein with the proper conformations of the protein induced. By imposing relatively soft conformational restraints to the protein during docking, this model reduced computational costs yet permitted essential conformational changes that were essential for these inhibitors to dock properly to the protein. For example, without adequate movement of the glycine-rich loop, it was difficult for the ligands to move from the surface of the protein to the binding site. In addition, these simulations called for better ways to compare simulation results with experiment other than using the popular root-mean-square deviation between the structure of a ligand in a docking pose and that in experiment because the structure of the protein also changed. In this work, we also calculated the correlation coefficient between protein-ligand/protein-protein distances in the docking structure and those in the crystal structure to check how well a ligand docked into the binding site of the protein and whether the proper conformation of the protein was induced.     

Flexible matching of test ligands to a 3D pharmacophore using a molecular superposition force field: Comparison of predicted and experimental conformations of inhibitors of three enzymes

Summary A computer procedure TFIT, which uses a molecular superposition force field to flexibly match test compounds to a 3D pharmacophore, was evaluated to find out whether it could reliably predict the bioactive conformations of flexible ligands. The program superposition force field optimizes the overlap of those atoms of the test ligand and template that are of similar chemical type, by applying an attractive force between atoms of the test ligand and template which are close together and of similar type (hydrogen bonding, charge, hydrophobicity). A procedure involving Monte Carlo torsion perturbations, followed by torsional energy minimization, is used to find conformations of the test ligand which cominimize the internal energy of the ligand and the superposition energy of ligand and template. The procedure was tested by applying it to a series of flexible ligands for which the bioactive conformation was known experimentally. The 15 molecules tested were inhibitors of thermolysin, HIV-1 protease or endothiapepsin for which X-ray structures of the bioactive conformation were available. For each enzyme, one of the molecules served as a template and the others, after being conformationally randomized, were fitted. The fitted conformation was then compared to the known binding geometry. The matching procedure was successful in predicting the bioactive conformations of many of the structures tested. Significant deviation from experimental results was found only for parts of molecules where it was readily apparent that the template did not contain sufficient information to accurately determine the bioactive conformation.     

Voronoi binding site models

A frequently occurring problem in drug design and enzymology is that the binding constants for several compounds to the same site are known, but the geometry and energetic interactions of the site are not. This paper presents in detail a novel approach to the problem which accurately but compactly represents the allowed conformation space of each ligand, accurately depicts their three-dimensional structures, and realistically allows each ligand to adopt the conformation and positioning in the site which is most favorable energetically. The investigator supplies only the ligand structures and observed binding free energies, along with a proposed site geometry. With no further assumptions about how the ligands bind and what parts of the ligands are important in determining the binding, the algorithm fits the observed binding energies without leaving outliers, predicts exactly how each of the given ligands binds in the site, and predicts the strength and mode of binding of new compounds, regardless of chemical similarity to the original set of ligands. The method is illustrated by devising a simple site that accounts for the binding of five polychlorinated biphenyls to thyroxine binding prealbumin. This model then predicts the binding energies correctly for an additional six biphenyls, and fails on one compound.     

An unconventional ligand-binding mechanism of substrate-binding proteins: MD simulation and Markov state model analysis of BtuF

In conventional “Venus Flytrap” mechanism, substrate-binding proteins (SBPs) interconvert between the open and closed conformations. Upon ligand binding, SBPs form a tightly closed conformation with the ligand bound at the interface of two domains. This mechanism was later challenged by many type III SBPs, such as the vitamin B₁₂-binding protein BtuF, in which the apo- and holo-state proteins adopt very similar conformations. Here, we combined molecular dynamics simulation and Markov state model analysis to study the conformational dynamics of apo- and B₁₂-bound BtuF. The results indicate that the crystal structures represent the only stable conformation of BtuF. Meanwhile, both apo- and holo-BtuF undergo large-scale interdomain motions with little energy cost. B₁₂ binding casts little restraints on the interdomain motions, suggesting that ligand binding affinity is enhanced by the remaining conformational entropy of holo-BtuF. These results reveal a new paradigm of ligand recognition mechanism of SBPs. © 2019 Wiley Periodicals, Inc.     

A pose prediction approach based on ligand 3D shape similarity

Molecular docking predicts the best pose of a ligand in the target protein binding site by sampling and scoring numerous conformations and orientations of the ligand. Failures in pose prediction are often due to either insufficient sampling or scoring function errors. To improve the accuracy of pose prediction by tackling the sampling problem, we have developed a method of pose prediction using shape similarity. It first places a ligand conformation of the highest 3D shape similarity with known crystal structure ligands into protein binding site and then refines the pose by repacking the side-chains and performing energy minimization with a Monte Carlo algorithm. We have assessed our method utilizing CSARdock 2012 and 2014 benchmark exercise datasets consisting of co-crystal structures from eight proteins. Our results revealed that ligand 3D shape similarity could substitute conformational and orientational sampling if at least one suitable co-crystal structure is available. Our method identified poses within 2 Å RMSD as the top-ranking pose for 85.7 % of the test cases. The median RMSD for our pose prediction method was found to be 0.81 Å and was better than methods performing extensive conformational and orientational sampling within target protein binding sites. Furthermore, our method was better than similar methods utilizing ligand 3D shape similarity for pose prediction.     

Approaches to solving the rigid receptor problem by identifying a minimal set of flexible residues during ligand docking

BACKGROUND: Using fixed receptor sites derived from high-resolution crystal structures in structure-based drug design does not properly account for ligand-induced enzyme conformational change and imparts a bias into the discovery and design of novel ligands. We sought to facilitate the design of improved drug leads by defining residues most likely to change conformation, and then defining a minimal manifold of possible conformations of a target site for drug design based on a small number of identified flexible residues. RESULTS: The crystal structure of thymidylate synthase from an important pathogenic target Pneumocystis carinii (PcTS) bound to its substrate and the inhibitor, BW1843U89, is reported here and reveals a new conformation with respect to the structure of PcTS bound to substrate and the more conventional antifolate inhibitor, CB3717. We developed an algorithm for determining which residues provide 'soft spots' in the protein, regions where conformational adaptation suggests possible modifications for a drug lead that may yield higher affinity. Remodeling the active site of thymidylate synthase with new conformations for only three residues that were identified with this algorithm yields scores for ligands that are compatible with experimental kinetic data. CONCLUSIONS: Based on the examination of many protein/ligand complexes, we develop an algorithm (SOFTSPOTS) for identifying regions of a protein target that are more likely to accommodate plastically to regions of a drug molecule. Using these indicators we develop a second algorithm (PLASTIC) that provides a minimal manifold of possible conformations of a protein target for drug design, reducing the bias in structure-based drug design imparted by structures of enzymes co-crystallized with inhibitors.     

Statistically based reduced representation of amino acid side chains

Preferred conformations of amino acid side chains have been well established through statistically obtained rotamer libraries. Typically, these provide bond torsion angles allowing a side chain to be traced atom by atom. In cases where it is desirable to reduce the complexity of a protein representation or prediction, fixing all side-chain atoms may prove unwieldy. Therefore, we introduce a general parametrization to allow positions of representative atoms (in the present study, these are terminal atoms) to be predicted directly given backbone atom coordinates. Using a large, culled data set of amino acid residues from high-resolution protein crystal structures, anywhere from 1 to 7 preferred conformations were observed for each terminal atom of the non-glycine residues. Side-chain length from the backbone C(alpha) is one of the parameters determined for each conformation, which should itself be useful. Prediction of terminal atoms was then carried out for a second, nonredundant set of protein structures to validate the data set. Using four simple probabilistic approaches, the Monte Carlo style prediction of terminal atom locations given only backbone coordinates produced an average root mean-square deviation (RMSD) of approximately 3 A from the experimentally determined terminal atom positions. With prediction using conditional probabilities based on the side-chain chi(1) rotamer, this average RMSD was improved to 1.74 A. The observed terminal atom conformations therefore provide reasonable and potentially highly accurate representations of side-chain conformation, offering a viable alternative to existing all-atom rotamers for any case where reduction in protein model complexity, or in the amount of data to be handled, is desired. One application of this representation with strong potential is the prediction of charge density in proteins. This would likely be especially valuable on protein surfaces, where side chains are much less likely to be fixed in single rotamers. Prediction of ensembles of structures provides a method to determine the probability density of charge and atom location; such a prediction is demonstrated graphically.     

Fully automated molecular mechanics based induced fit protein-ligand docking method

We describe a method for docking a ligand into a protein receptor while allowing flexibility of the protein binding site. The method employs a multistep procedure that begins with the generation of protein and ligand conformations. An initial placement of the ligand is then performed by computing binding site hotspots. This initial placement is followed by a protein side-chain refinement stage that models protein flexibility. The final step of the process is an energy minimization of the ligand pose in the presence of the rigid receptor. Thus the algorithm models flexibility of the protein at two stages, before and after ligand placement. We validated this method by performing docking and cross docking studies of eight protein systems for which crystal structures were available for at least two bound ligands. The resulting rmsd values of the 21 docked protein-ligand complexes showed values of 2 A or less for all but one of the systems examined. The method has two critical benefits for high throughput virtual screening studies. First, no user intervention is required in the docking once the initial binding site selection has been made in the protein. Second, the initial protein conformation generation needs to be performed only once for a given binding region. Also, the method may be customized in various ways depending on the particular scenario in which dockings are being performed. Each of the individual steps of the method is fully independent making it straightforward to explore different variants of the high level workflow to further improve accuracy and performance.