AsteriX: a Web server to automatically extract ligand coordinates from figures in PDF articles.

Coordinates describing the chemical structures of small molecules that are potential ligands for pharmaceutical targets are used at many stages of the drug design process. The coordinates of the vast majority of ligands can be obtained from either publicly accessible or commercial databases. However, interesting ligands sometimes are only available from the scientific literature, in which case their coordinates need to be reconstructed manually--a process that consists of a series of time-consuming steps. We present a Web server that helps reconstruct the three-dimensional (3D) coordinates of ligands for which a two-dimensional (2D) picture is available in a PDF file. The software, called AsteriX, analyses every picture contained in the PDF file and attempts to determine automatically whether or not it contains ligands. Areas in pictures that may contain molecular structures are processed to extract connectivity and atom type information that allow coordinates to be subsequently reconstructed. The AsteriX Web server was tested on a series of articles containing a large diversity in graphical representations. In total, 88% of 3249 ligand structures present in the test set were identified as chemical diagrams. Of these, about half were interpreted correctly as 3D structures, and a further one-third required only minor manual corrections. It is principally impossible to always correctly reconstruct 3D coordinates from pictures because there are many different protocols for drawing a 2D image of a ligand, but more importantly a wide variety of semantic annotations are possible. The AsteriX Web server therefore includes facilities that allow the users to augment partial or partially correct 3D reconstructions. All 3D reconstructions are submitted, checked, and corrected by the users domain at the server and are freely available for everybody. The coordinates of the reconstructed ligands are made available in a series of formats commonly used in drug design research. The AsteriX Web server is freely available at http://swift.cmbi.ru.nl/bitmapb/.

Drug design for ever, from hype to hope.

In its first 25 years JCAMD has been disseminating a large number of techniques aimed at finding better medicines faster. These include genetic algorithms, COMFA, QSAR, structure based techniques, homology modelling, high throughput screening, combichem, and dozens more that were a hype in their time and that now are just a useful addition to the drug-designers toolbox. Despite massive efforts throughout academic and industrial drug design research departments, the number of FDA-approved new molecular entities per year stagnates, and the pharmaceutical industry is reorganising accordingly. The recent spate of industrial consolidations and the concomitant move towards outsourcing of research activities requires better integration of all activities along the chain from bench to bedside. The next 25 years will undoubtedly show a series of translational science activities that are aimed at a better communication between all parties involved, from quantum chemistry to bedside and from academia to industry. This will above all include understanding the underlying biological problem and optimal use of all available data.

How do substrates enter and products exit the buried active site of cytochrome P450cam? 1. Random expulsion molecular dynamics investigation of ligand access channels and mechanisms.

Cytochrome P450s form a ubiquitous protein family with functions including the synthesis and degradation of many physiologically important compounds and the degradation of xenobiotics. Cytochrome P450cam from Pseudomonas putida has provided a paradigm for the structural understanding of cytochrome P450s. However, the mechanism by which camphor, the natural substrate of cytochrome P450cam, accesses the buried active site is a long-standing puzzle. While there is recent crystallographic and simulation evidence for opening of a substrate-access channel in cytochrome P450BM-3, for cytochrome P450cam, no such conformational changes have been observed either in different crystal structures or by standard molecular dynamics simulations. Here, a novel simulation method, random expulsion molecular dynamics, is presented, in which substrate-exit channels from the buried active site are found by imposing an artificial randomly oriented force on the substrate, in addition to the standard molecular dynamics force field. The random expulsion molecular dynamics method was tested in simulations of the substrate-bound structure of cytochrome P450BM-3, and then applied to complexes of cytochrome P450cam with different substrates and with product. Three pathways were identified, one of which corresponds to a channel proposed earlier on the basis of crystallographic and site-directed mutagenesis data. Exit via the water-filled channel, which was previously suggested to be a product exit channel, was not observed. The pathways obtained by the random expulsion molecular dynamics method match well with thermal motion pathways obtained by an analysis of crystallographic B-factors. In contrast to large backbone motions (up to 4 A) observed in cytochrome P450BM-3 for the exit of palmitoleic acid, passage of camphor through cytochrome P450cam only requires small backbone motions (less than 2.4 A) in conjunction with side-chain rotations. Concomitantly, in almost all the exit trajectories, salt-links that have been proposed to act as ionic tethers between secondary structure elements of the protein, are perturbed.

How do substrates enter and products exit the buried active site of cytochrome P450cam? 2. Steered molecular dynamics and adiabatic mapping of substrate pathways.

Three possible channels by which substrates and products can exit from the buried active site of cytochrome P450cam have been identified by means of random expulsion molecular dynamics simulations. In the investigation described here, we computed estimates of the relative probabilities of ligand passage through the three channels using steered molecular dynamics and adiabatic mapping. For comparison, the same techniques are also applied to investigate substrate egress from cytochrome P450-BM3. The channel in cytochrome P450cam, for which there is the most supporting evidence from experiments (which we name pathway 2a), is computed to be the most probable ligand exit channel. It has the smallest computed unbinding work and force. For this channel, the ligand exits between the F/G loop and the B' helix. Two mechanistically distinct, but energetically similar routes through this channel were observed, showing that multiple pathways along one channel are possible. The probability of ligand exit via the next most probable channel (pathway 3), which is located between the I helix and the F and G helices, is estimated to be less than 1/10 of the probability of exit along pathway 2a. Low-frequency modes of the protein extracted from an essential dynamics analysis of a 1 ns duration molecular dynamics simulation of cytochrome P450cam with camphor bound, support the opening of pathway 2a on a longer timescale. On longer timescales, it is therefore expected that this pathway becomes more dominant than estimated from the present computations.

Multicopy molecular dynamics simulations suggest how to reconcile crystallographic and product formation data for camphor enantiomers bound to cytochrome P-450cam.

Multiple ligand binding modes are possible in many enzyme active sites; their presence in cytochrome P450cam (P450cam) is evident from crystallographic studies of the binding of thiocamphor and phenylimidazoles. Here, we use multicopy molecular dynamics simulations to compare the binding modes of (1R)- and (1S)-camphor in the active site of P450cam. Simulations with (1R)-camphor, the natural substrate, serve to calibrate our protocol: 19 out of 20 copies of (1R)-camphor converged to coordinates very close to those observed for (1R)-camphor in its crystallographic complex with P450cam during the simulations. Simulations with the (1S)-camphor enantiomer showed greater mobility of the substrate, consistent with spectroscopic data, and resulted in 3 major binding modes. One of these is similar to the major conformation (of the two conformations assigned) in a recently determined crystal structure, but this conformation is not correctly oriented for regiospecific hydroxylation at C-5. The simulations, however, provide evidence for reorientation of (1S)-camphor upon formation of the reactive Fe-O intermediate to an orientation suitable for hydroxylation. The simulations thus permit rationalisation of the apparent inconsistency between the crystal structure and the reaction products.

Towards molecular dynamics simulation of large proteins with a hydration shell at constant pressure.

Molecular dynamics simulation of a large protein in explicit water with periodic boundary conditions is extremely demanding in terms of computation time. Consequently, we have sought approximations of the solvent environment that model its important features. Here, we describe our SAPHYR (Shell Approximation for Protein HYdRation) model in which the protein is surrounded by a shell of water molecules maintained at constant pressure. In addition to the usual pairwise interatomic interactions, these water molecules are subjected to forces approximating van der Waals and dipole-dipole interactions with the implicit surrounding bulk solvent. The SAPHYR model is tested for a system of one argon atom in water and for the protein ubiquitin, and then applied to cytochrome P450cam, a protein with over 400 residues. The results demonstrate that structural and dynamic properties of the simulated systems are improved by use of the SAPHYR model, and that this model provides a significant computational saving over simulations with periodic boundary conditions.

pKa calculations for class A beta-lactamases: influence of substrate binding.

Beta-Lactamases are responsible for bacterial resistance to beta-lactams and are thus of major clinical importance. However, the identity of the general base involved in their mechanism of action is still unclear. Two candidate residues, Glu166 and Lys73, have been proposed to fulfill this role. Previous studies support the proposal that Glu166 acts during the deacylation, but there is no consensus on the possible role of this residue in the acylation step. Recent experimental data and theoretical considerations indicate that Lys73 is protonated in the free beta-lactamases, showing that this residue is unlikely to act as a proton abstractor. On the other hand, it has been proposed that the pKa of Lys73 would be dramatically reduced upon substrate binding and would thus be able to act as a base. To check this hypothesis, we performed continuum electrostatic calculations for five wild-type and three beta-lactamase mutants to estimate the pKa of Lys73 in the presence of substrates, both in the Henri-Michaelis complex and in the tetrahedral intermediate. In all cases, the pKa of Lys73 was computed to be above 10, showing that it is unlikely to act as a proton abstractor, even when a beta-lactam substrate is bound in the enzyme active site. The pKa of Lys234 is also raised in the tetrahedral intermediate, thus confirming a probable role of this residue in the stabilization of the tetrahedral intermediate. The influence of the beta-lactam carboxylate on the pKa values of the active-site lysines is also discussed.

Comparative kinetic analysis of FLP and cre recombinases: mathematical models for DNA binding and recombination.

The integrase class site specific recombinases FLP from Saccharomyces cerevisiae, and Cre from bacteriophage P1, have been extensively used to direct DNA rearrangements in heterologous organisms. Although their reaction mechanisms have been relatively well characterised, little comparative analysis of the two enzymes has been published. We present a comparative kinetic analysis of FLP and Cre, which identifies important differences. Gel mobility shift assays show that Cre has a higher affinity for its target, loxP (7. 4x10(10) M-1), than FLP for its target, FRT (8.92x10(8) M-1). We show that both recombinases bind the two halves of their target sites cooperatively, and that Cre shows approximately threefold higher cooperativity than FLP. Using a mathematical model describing the sequential binding of recombinase monomers to DNA, we have determined values for the association and dissociation rate constants for FLP and Cre.FLP and Cre also showed different characteristics in in vitro recombination assays. In particular, approximately tenfold more active FLP was required than Cre to optimally recombine a given quantity of excision substrate. FLP was able to reach maximum excision levels approaching 100%, whilst Cre-mediated excision did not exceed 75%. To investigate possible reasons for these differences a mathematical model describing the excision recombination reaction was established. Using measured DNA binding parameters for FLP and Cre in the model, and comparing simulated and experimental recombination data, the values of the remaining unknown parameters were determined. This analysis indicates that the synaptic complex is more stable for Cre than for FLP.

Electrostatic steering and ionic tethering in enzyme-ligand binding: insights from simulations.

To bind at an enzyme's active site, a ligand must diffuse or be transported to the enzyme's surface, and, if the binding site is buried, the ligand must diffuse through the protein to reach it. Although the driving force for ligand binding is often ascribed to the hydrophobic effect, electrostatic interactions also influence the binding process of both charged and nonpolar ligands. First, electrostatic steering of charged substrates into enzyme active sites is discussed. This is of particular relevance for diffusion-influenced enzymes. By comparing the results of Brownian dynamics simulations and electrostatic potential similarity analysis for triose-phosphate isomerases, superoxide dismutases, and beta-lactamases from different species, we identify the conserved features responsible for the electrostatic substrate-steering fields. The conserved potentials are localized at the active sites and are the primary determinants of the bimolecular association rates. Then we focus on a more subtle effect, which we will refer to as "ionic tethering." We explore, by means of molecular and Brownian dynamics simulations and electrostatic continuum calculations, how salt links can act as tethers between structural elements of an enzyme that undergo conformational change upon substrate binding, and thereby regulate or modulate substrate binding. This is illustrated for the lipase and cytochrome P450 enzymes. Ionic tethering can provide a control mechanism for substrate binding that is sensitive to the electrostatic properties of the enzyme's surroundings even when the substrate is nonpolar.

pKa calculations for class A beta-lactamases: methodological and mechanistic implications.

Beta-lactamases are responsible for resistance to penicillins and related beta-lactam compounds. Despite numerous studies, the identity of the general base involved in the acylation step is still unclear. It has been proposed, on the basis of a previous pKa calculation and analysis of structural data, that the unprotonated Lys73 in the active site could act as the general base. Using a continuum electrostatic model with an improved treatment of the multiple titration site problem, we calculated the pKa values of all titratable residues in the substrate-free TEM-1 and Bacillus licheniformis class A beta-lactamases. The pKa of Lys73 in both enzymes was computed to be above 10, in good agreement with recent experimental data on the TEM-1 beta-lactamase, but inconsistent with the proposal that Lys73 acts as the general base. Even when the closest titratable residue, Glu166, is mutated to a neutral residue, the predicted downward shift of the pKa of Lys73 shows that it is unlikely to act as a proton abstractor in either enzyme. These results support a mechanism in which the proton of the active Ser70 is transferred to the carboxylate group of Glu166.

Exceptionally stable salt bridges in cytochrome P450cam have functional roles.

A long-standing puzzle in structure-function studies of cytochrome P450cam is how the substrate, camphor, reaches the buried active site. The crystal structure shows no channel from the surface to the active site large enough for substrate to pass through. Recent experiments indicate that access of the rather nonpolar substrate to the active site is controlled by electrostatic interactions and may involve rupture of the two salt links to Asp251 [Deprez, E., Gerber, N. C., Di Primo, C., Douzou, P., Sligar, S. G., & Hui Bon Hoa, G. (1994) Biochemistry 33, 14464-14468]. Consequently, we have computed the electrostatic strength of 53 ionic pairs, including 32 salt links, in cytochrome P450cam by numerical solution of the finite-difference linearized Poisson-Boltzmann equation. The calculated electrostatic free energies, delta Gtot, of the salt links range from -9 to +6 kcal/mol with approximately 60% of the salt links being energetically favorable and 40% being unfavorable with respect to mutation to their uncharged, nonpolar isosteres. Strikingly, of the four most stable salt links in the protein (delta Gtot < -6 kcal/mol), two involve the propionate groups of the heme and the other two involve Asp251. In the modeled D251N mutant, for which electrostatic effects on substrate binding are diminished, the latter two salt links lose their stability (delta Gtot > -2.4 kcal/mol). Thus it appears that cytochrome P450cam has evolved four unusually strong salt bridges, stabilized by surrounding charged and polar groups in the protein, to keep its heme cofactor in place and to regulate substrate binding.

A global model of the protein-solvent interface.

The solvent structure and dynamics around myoglobin is investigated at the microscopic level of detail by computer simulation. We analyze a molecular dynamics trajectory in terms of solvent mobility and probability distribution. Local events, occurring in the protein-solvent interfacial region, which are often masked by other approaches are thus revealed. Specifically, the local solvent mobility is greatly enhanced for certain locations at the protein surface and in its interior. In addition, a strong correlation between the solvent mobility and density emerges on both global and local scales. We propose a simple model where the solvent distribution measured perpendicularly to the protein surface is utilized to reconstruct the simulated network of hydration within 6 A from the protein surface with a relative error of only 17%. The global precision of this solvation model matches results obtained with more complicated models usually used in refinement procedures in x-ray and neutron experiments but with far fewer parameters. The dramatically improved correspondence between observed and calculated x-ray intensities at low resolution relative to other methods both confirms the validity of the approach used in the MD (molecular dynamics) simulations and allows the results of this study to be implemented in solvent studies on real systems.

A connected-cluster of hydration around myoglobin: correlation between molecular dynamics simulations and experiment.

An analysis of a molecular dynamics simulation of metmyoglobin in an explicit solvent environment of 3,128 water molecules has been performed. Both statics and dynamics of the protein-solvent interface are addressed in a comparison with experiment. Three-dimensional density distributions, temperature factors, and occupancy weights are computed for the solvent by using the trajectory coordinates. Analysis of the hydration leads to the localization of more than 500 hydration sites distributed into multiple layers of solvation located between 2.6 and 6.8 A from the atomic protein surface. After locating the local solvent density maxima or hydration sites we conclude that water molecules of hydration positions and hydration sites are distinct concepts. Both global and detailed properties of the hydration cluster around myoglobin are compared with recent neutron and X-ray data on myoglobin. Questions arising from differences between X-ray and neutron data concerning the locations of the protein-bound water are investigated. Analysis of water site differences found from X-ray and neutron experiments compared with our simulation shows that the simulation gives a way to unify the hydration picture given by the two experiments.

Distribution function implied dynamics versus residence times and correlations: solvation shells of myoglobin.

The dynamics of water at the protein-solvent interface is investigated through the analysis of a molecular dynamics simulation of metmyoglobin in explicit aqueous environment. Distribution implied dynamics, harmonic and quasi-harmonic, are compared with the simulated macroscopic dynamics. The distinction between distinguishable solvent molecules and hydration sites developed in the previous paper is used. The simulated hydration region within 7 A from the protein surface is analyzed using a set of 551 hydration sites characterized by occupancy weights and temperature B-factors determined from the simulation trajectory. The precision of the isotropic harmonic and anisotropic harmonic models for the description of proximal solvent fluctuations is examined. Residence times and dipole reorientation times of water around the protein surface are compared with NMR and ESR results. A correlation between diffraction experiment quantities such as the occupancy weights and temperature factors and the residence and correlation times resulting from magnetic resonance experiments is found via comparison with simulation.