Prediction of Properties from Simulations: Free Energies of Solvation in Hexadecane, Octanol, and Water

Calculation of the aqueous solvation energy and entropy, as well as free energy, of simple polar solutes

With the advent of more powerful computers, the question of calculating thermodynamic quantities, such as the energy and the entropy, in solute-solvent systems is revisited. The calculation of these thermodynamic quantitites was limited in the past by their slow convergence relative to the free energy. Using molecular dynamics simulations, the energy, entropy, and free energy of solvation of NMA and CH(3)NH(2), as well as their relative values, have been determined. Three different methods (the thermodynamic perturbation method, the thermodynamic integration method, and a finite-difference method) are compared. The finite difference method gives the best results and accurate values for the entropy and energy were obtained using a reasonable amount to computer time. The results suggest that a meaningful thermodynamic description of biomolecular processes can be realized with present methods and the available computer time.

Note on the energy density in the solvent induced by a solute

The question of how far the effect of the presence of a solute molecule propagates into the solvent is studied in a lattice model that had been used earlier to describe hydrophobic interactions. The local energy density in the model solvent is obtained as an explicit function of distance from the solute and is found to decay to its bulk-phase value with the same decay length as that of the solvent-mediated part of the potential of mean force between a pair of solute molecules. The integrated deviation of the energy density from its bulk-phase value is evaluated in Bethe-Guggenheim approximation and shown to be identical to the energy change accompanying the dissolution of the solute as obtained from the temperature dependence of its solubility.

Predicting Physical−Chemical Properties of Compounds from Molecular Structures by Recursive Neural Networks

In this paper, we report on the potential of a recently developed neural network for structures applied to the prediction of physical chemical properties of compounds. The proposed recursive neural network (RecNN) model is able to directly take as input a structured representation of the molecule and to model a direct and adaptive relationship between the molecular structure and target property. Therefore, it combines in a learning system the flexibility and general advantages of a neural network model with the representational power of a structured domain. As a result, a completely new approach to quantitative structure-activity relationship/quantitative structure-property relationship (QSPR/QSAR) analysis is obtained. An original representation of the molecular structures has been developed accounting for both the occurrence of specific atoms/groups and the topological relationships among them. Gibbs free energy of solvation in water, Delta(solv)G degrees , has been chosen as a benchmark for the model. The different approaches proposed in the literature for the prediction of this property have been reconsidered from a general perspective. The advantages of RecNN as a suitable tool for the automatization of fundamental parts of the QSPR/QSAR analysis have been highlighted. The RecNN model has been applied to the analysis of the Delta(solv)G degrees in water of 138 monofunctional acyclic organic compounds and tested on an external data set of 33 compounds. As a result of the statistical analysis, we obtained, for the predictive accuracy estimated on the test set, correlation coefficient R = 0.9985, standard deviation S = 0.68 kJ mol(-1), and mean absolute error MAE = 0.46 kJ mol(-1). The inherent ability of RecNN to abstract chemical knowledge through the adaptive learning process has been investigated by principal components analysis of the internal representations computed by the network. It has been found that the model recognizes the chemical compounds on the basis of a nontrivial combination of their chemical structure and target property.

Insight into the solvent effect: A density functional theory study of cisplatin hydrolysis

The solvent effect on reactions in solutions is crucial for many systems. In this study, the reaction barrier with respect to the number of solvent molecules included in the system is systematically studied using density function theory calculations. Our results show that the barriers rapidly converge with respect to the number of solvent molecules. The solvent effect is investigated by calculating cisplatin hydrolysis in several types of solvents. The results are analyzed and a linear relationship between the reaction barrier and the interaction strength of solvent-reactants is found. Insight into the general solvent effect is obtained.

Theoretical Study on the Properties of Linear and Cyclic Amides in Gas Phase and Water Solution

The structural and energetic properties of a group of selected amides, of well-known importance for the design of efficient clathrate inhibitors, are calculated with Hartree-Fock and density functional theory, B3LYP, theoretical levels, and a 6-311++g** basis set in the gas phase and a water solution. The conformational behavior of the molecules is studied through the scanning of the torsional potential energy surfaces and by the analysis of the differences in the energetic and structural properties between the isomers. The properties of the amides in water solution are determined within a self-consistent reaction field approach with a polarizable continuum model that allows the calculation of the different contributions to the free energy of solvation. The calculated barriers to rotation are in good agreement with the available experimental data and the comparison of the gas and water results shows the strong effect of the solute polarization. The properties of different amide-water complexes are calculated and compared with available experimental information.

Solution-phase synthesis of a tricyclic pyrrole-2-carboxamide discovery library applying a stetter-Paal-Knorr reaction sequence

The solution-phase synthesis of a discovery library of 178 tricyclic pyrrole-2-carboxamides was accomplished in nine steps and seven purifications starting with three benzoyl-protected amino acid methyl esters. Further diversity was introduced by two glyoxaldehydes and 41 primary amines. The combination of Pauson-Khand, Stetter, and microwave-assisted Paal-Knorr reactions was applied as a key sequence. The discovery library was designed with the help of QikProp 2.1, and physicochemical data are presented for all pyrroles. Library members were synthesized and purified in parallel and analyzed by LC/MS. Selected compounds were fully characterized.

Structure-based drug design meets the ribosome

The high-resolution structures of the bacterial ribosomal subunits and those of their complexes with antibiotics have advanced significantly our understanding of small-molecule interactions with RNA. The wealth of RNA structural data generated by these structures has allowed computational chemists to employ a drug discovery paradigm focused on RNA-based targets. The structures also show how target-based resistance affects antibiotics acting at the level of the ribosome. Not only are the sites pinpointed where different classes of antibiotics inhibit protein synthesis, but their orientations, relative dispositions, and unique mechanisms of action are also revealed at the atomic level. Both the 30S and the 50S ribosomal subunits have been shown to be "targets of targets", offering several adjacent, functionally relevant binding pockets for antibiotics. It is the detailed knowledge of these validated locations, or ribofunctional loci, plus the mapping of the resistance hot-spots that allow the rational design of next-generation antibacterials. When the structural information is combined with a data-driven computational toolkit able to describe and predict molecular properties appropriate for bacterial cell penetration and drug-likeness, a structure-based drug design approach for novel antibacterials shows great promise.

Development and Evaluation of an in Silico Model for hERG Binding

It has been recognized that drug-induced QT prolongation is related to blockage of the human ether-a-go-go-related gene (hERG) ion channel. Therefore, it is prudent to evaluate the hERG binding of active compounds in early stages of drug discovery. In silico approaches provide an economic and quick method to screen for potential hERG liability. A diverse set of 90 compounds with hERG IC(50) inhibition data was collected from literature references. Fragment-based QSAR descriptors and three different statistical methods, support vector regression, partial least squares, and random forests, were employed to construct QSAR models for hERG binding affinity. Important fragment descriptors relevant to hERG binding affinity were identified through an efficient feature selection method based on sparse linear support vector regression. The support vector regression predictive model built upon selected fragment descriptors outperforms the other two statistical methods in this study, resulting in an r(2) of 0.912 and 0.848 for the training and testing data sets, respectively. The support vector regression model was applied to predict hERG binding affinities of 20 in-house compounds belonging to three different series. The model predicted the relative binding affinity well for two out of three compound series. The hierarchical clustering and dendrogram results show that the compound series with the best prediction has much higher structural similarity and more neighbors of training compounds than the other two compound series, demonstrating the predictive scope of the model. The combination of a QSAR model and postprocessing analysis, such as clustering and visualization, provides a way to assess the confidence level of QSAR prediction results on the basis of similarity to the training set.

A combination of docking, QM/MM methods, and MD simulation for binding affinity estimation of metalloprotein ligands

To alleviate the problems in the receptor-based design of metalloprotein ligands due to inadequacies in the force-field description of coordination bonds, a four-tier approach was devised. Representative ligand-metalloprotein interaction energies are obtained by subsequent application of (1) docking with metal-binding-guided selection of modes, (2) optimization of the ligand-metalloprotein complex geometry by combined quantum mechanics and molecular mechanics (QM/MM) methods, (3) conformational sampling of the complex with constrained metal bonds by force-field-based molecular dynamics (MD), and (4) a single point QM/MM energy calculation for the time-averaged structures. The QM/MM interaction energies are, in a linear combination with the desolvation-characterizing changes in the solvent-accessible surface areas, correlated with experimental data. The approach was applied to structural correlation of published binding free energies of a diverse set of 28 hydroxamate inhibitors to zinc-dependent matrix metalloproteinase 9 (MMP-9). Inclusion of steps 3 and 4 significantly improved both correlation and prediction. The two descriptors explained 90% of variance in inhibition constants of all 28 inhibitors, ranging from 0.08 to 349 nM, with the average unassigned error of 0.318 log units. The structural and energetic information obtained from the time-averaged MD simulation results helped understand the differences in binding modes of related compounds.

Prediction of Properties from Simulations: A Re-examination with Modern Statistical Methods

We discuss models fit to data collected by Duffy and Jorgensen to predict solvation free energies and partition equilibria of drugs, organic molecules, aromatic heterocycles, and other molecules. These data were originally examined using linear regression, but here more recently developed statistical models are applied. The data set is complicated due to the presence of discrepant observations and also curvature in the response. In some cases it is possible to discard a small number of the observations to get good fit to the data, but, in others, discarding an increasing proportion of the observations does not improve the fit. Our general preference is to use robust parameter estimation which downweights to reduce the influence of discrepant observations on the fitted models. Models are selected for four responses using linear or more complicated representations of the explanatory variables, such as cubic polynomials, B-splines, or smoothers via generalized additive models (GAMs). Variables are chosen using the traditional approach of formal tests to assess contribution to the fit of a model, and resampling methods including bootstrap are also considered to assess the prediction error for given models. Results of our analysis indicate that GAMs are an improvement on linear models for describing the data and making predictions. In general robust regression models and GAMs have the smallest conditional expected loss of prediction over the four responses. In addition, robust regression models offer the advantage of identifying molecules that perform poorly in the fit. In general, models were identified that yielded an improvement of approximately 50% in the conditional expected loss of prediction compared with the original parametrization of Duffy and Jorgensen. It was also found that the use of cross-validation to compare models was unreliable, and bootstrapping is preferred.

Amine Hydration: A United-Atom Force-Field Solution

The free energies of hydration for ammonia and mono-, di-, and trimethylated amines experimentally show an unexpected trend that has, in the past, been difficult to reproduce computationally. Absolute and relative free energies of hydration of these compounds were calculated using the OPLS all-atom and the united-atom GROMOS force fields. Both force fields reproduce the relative free energy of hydration, but the absolute free energies of hydration were only reproduced within kBT when using the recently developed GROMOS parameter set 53A6. Relative free energies of solvation in chloroform could also be reproduced indicating a proper partitioning of the compounds between polar and apolar media. Overall we conclude that it is possible to resolve the amine hydration problem using a simple united-atom force field.

Molecular Dynamics Simulation of Water Near Nanostructured Hydrophobic Surfaces: Interfacial Energies

We present results from molecular dynamics simulations of water near structured hydrophobic surfaces. The surface structures reported herein are a planar alkane crystal as a reference and crystals with a hole and a protrusion of approximately 2.5 nm diameter and 0.5 nm depth or height. All indicators show that surface structuring increases the hydrophobicity: The water density is reduced near the structure elements, and the number of residual contacts between water and the surface decreases by about 40 % with respect to the planar surface. Thermodynamic integration shows that the interfacial energy of the structured surfaces is about 7 mJ m(-2) higher for structured surfaces than for the planar surface. The hydrophobicity increases by a similar amount for the hole and the protrusion geometries compared to the planar surface.

In vitro and in vivo antimalarial activity of peptidomimetic protein farnesyltransferase inhibitors with improved membrane permeability

A series of protein farnesyltransferase inhibitor ester prodrugs of FTI-2148 (17) were synthesized in order to evaluate the effects of ester structure modification on antimalarial activity and for further development of a farnesyltransferase inhibitor with in vivo activity. Evaluation against P. falciparum in red blood cells showed that all the investigated esters exhibited significant antimalarial activity, with the benzyl ester 16 showing the best inhibition (ED50=150 nM). Additionally, compound 16 displayed in vivo activity and was found to suppress parasitemia by 46.1% at a dose of 50 mg kg(-1) day(-1) against Plasmodium berghei in mice. The enhanced inhibition potency of the esters is consistent with improved cell membrane permeability compared to that of the free acid. The results of this study suggest that protein farnesyltransferase is a valid antimalarial drug target and that the antimalarial activity of these compounds derives from a balance between the hydrophobic character and the size and conformation of the ester moiety.

Effect of nanostructure on the properties of water at the water-hydrophobic interface: a molecular dynamics simulation

The local structure of water near hydrophobic surfaces of different surface topographies has been analyzed by molecular dynamics simulation. An alkane crystal has been taken as the parent model for a hydrophobic surface. Surface structures were created by placing pits into it, which were half a nanometer deep and several nanometers wide. Around all structures, the water has a lower density, less orientational ordering, fewer water-water hydrogen bonds, and fewer surface contacts than for a flat unstructured surface. This indicates that the structured surfaces are more hydrophobic than the flat surface. Of the structures investigated, pits with a diameter of approximately 2.5 nm were effective in increasing the hydrophobic character of the surface.

Molecular Dynamics Simulation of Aqueous NaF and NaI Solutions near a Hydrophobic Surface

We present results from molecular dynamics simulation of aqueous solutions of alkali halide salts (NaI and NaF) at the interface with hydrophobic objects. The primary objective of this study is to investigate the structural properties of the salt solutions at the hydrophobic surface. An alkane crystal has been taken as the parent model for a hydrophobic surface. A hexagonal hole was created on it, which was half a nm deep and 2.5 nm wide. The density distributions of different species (water, anions, and cations) are studied as a function of distance from the surface. While iodide prefers the interface, the fluoride ions stay inside the bulk water region. The higher concentration of iodide ions at the interface drags sodium counterions to the interface. It also decreases the water density at the interface because of steric effects of the iodide ions. The number of contacts between the surface carbons and water decreases in the case of NaI solutions but is unchanged for NaF solutions. The orientation of the water-ion and the water-water hydrogen bond vector orientations near the interface is discussed in detail.

Protegrin structure-activity relationships: using homology models of synthetic sequences to determine structural characteristics important for activity

The protegrin family of antimicrobial peptides is among the shortest in sequence length while remaining very active against a variety of microorganisms. The major goal of this study is to characterize easily calculated molecular properties, which quantitatively show high correlation with antibacterial activity. The peptides studied have high sequence similarity but vary in activity over more than an order of magnitude. Hence, sequence analysis alone cannot be used to predict activity for these peptides. We calculate structural properties of 62 protegrin and protegrin-analogue peptides and correlate them to experimental activities against six microbe species, as well as hemolytic and cytotoxic activities. Natural protegrins structures were compared with synthetic derivatives using homology modeling, and property descriptors were calculated to determine the characteristics that confer their antimicrobial activity. A structure-activity relationship study of all these peptides provides information about the structural properties that affect activity against different microbial species.

The hydrophobicity of nanostructured alkane and perfluoro alkane surfaces: A comparison by molecular dynamics simulation

In this contribution we investigate the differences in the hydrophobicity of a perfluoro-n-eicosane crystal and the n-eicosane crystal by molecular dynamics simulation. The results were analysed in terms of density of water at the interface, the chemical potential of water at the interface and the orientational ordering of water at the interface. The perfluoro-n-eicosane crystal-water interface is found to have a less density, higher chemical potential and a weaker orientational ordering at the interface than the corresponding n-eicosane crystal.

Analyses of the partition coefficient, log P, using ab initio MO parameter and accessible surface area of solute molecules

To analyze the log P(sol/w) values (sol: n-octanol or chloroform, w: water) in the framework of the molecular orbital (MO) procedure, we selected solute descriptors such as the solvation energy difference between aqueous and organic solvent phases and the "surface" area of solute molecules to which water molecules are accessible. The solvation energy of solute molecules in their minimum free-energy conformation was calculated using the ab initio self-consistent reaction field-MO method with the conductor-like screening model. The experimentally measured log P(sol/w) value of various solutes except for those of amphiprotics was shown to be analyzable reasonably well by the MO model with additional descriptors for the hydrogen-bonding patterns in the solute-solvent interactions.

Theoretical Scales of Hydrogen Bond Acidity and Basicity for Application in QSAR/QSPR Studies and Drug Design. Partitioning of Aliphatic Compounds

Phenomenological analysis of existing hydrogen bond (HB) donor and acceptor scales and apparent physical considerations have enabled the establishment of new quantitative scales of hydrogen bond basicity and acidity. Chemical structures represented by molecular graphs and the orbital electronegativities of Hinze and Jaffe are utilized as an input data. The scales obtained correlate well with several experimental solvent polarity scales such as and, pK(HB), and E(T)(30). To demonstrate the applicability of the new quantities, we have applied them to seven equilibrium partitioning data sets: octanol-water, hexadecane-water, chloroform-water, gas-water, gas-octanol, gas-hexadecane, and gas-chloroform partition coefficients. The hydrogen bond descriptors when supplemented by a cavity-forming term and a dipolarity term show high performance in correlations of the partition coefficients of aliphatic compounds. These new HB descriptors can be used in studying hydrogen bonding and fluid phase equilibria as well as scoring functions in ligand docking and descriptors in ADME evaluations.

Theoretical Prediction of Partition Coefficients via Molecular Electrostatic and Electronic Properties

Previously published methods for calculation of Abraham's polarity/polarizability and hydrogen bond acidity and basicity descriptors are validated for their ability to predict the various partition coefficients of 80 challenging molecules. As well as this indirect validation, accurate log P predictions are shown to be possible by using directly the fundamental molecular properties used in the calculation of descriptors. From a general point of view, the van der Waals and hydrogen bond interactions present between the solute and the water/solvent system can be represented by charge-based interactions, which are partitioned into a positive and a negative term (Sigma(T)V(Max) and Sigma(T)V(Min)), dipole-based interactions (mu) and induced dipole-based interactions (alpha); further stabilization is possible if solute and solvent densities come into contact and overlap (Sigma(T)E). A discussion is opened on the possibility to extend this set to describe systems with electron donor/acceptor interactions other than hydrogen bonding.

Development and Use of Hydrophobic Surface Area (HSA) Descriptors for Computer-Assisted Quantitative Structure−Activity and Structure−Property Relationship Studies

A new series of 25 whole-molecule molecular structure descriptors are proposed. The new descriptors are termed Hydrophobic Surface Area, or HSA descriptors, and are designed to capture information regarding the structural features responsible for hydrophobic and hydrophilic intermolecular interactions. The utility of the HSAs in capturing this type of information is demonstrated using two properties that have a known hydrophobic component. The first study involves the modeling of the inhibition of Gram-positive bacteria cell growth of a series of biarylamides. The second application involves the study of the blood-brain barrier penetration of a diverse series of drug molecules. In both cases, the HSAs are shown to effectively capture information related to the hydrophobic components of these two properties. Additional evaluation of the new class of descriptors shows them to be unique in their ability to measure hydrophobic features among a diverse set of conventional structural descriptors. The HSAs are evaluated regarding their sensitivity to conformational changes and are found to be similar in that regard to other widely used molecular descriptors.

QSAR Studies of PC-3 cell line inhibition activity of TSA and SAHA-like hydroxamic acids

Quantitative structure-activity relationships (QSAR) for a series of new trichostatin A (TSA)-like hydroxamic acids for the inhibition of cell proliferation of the PC-3 cell line have been developed using molecular descriptors from Qikprop and electronic structure calculations. The best regression model shows that the PM3 atomic charge on the carbonyl carbon in the CONHOH moiety(Qco), globularity (Glob), and the hydrophilic component of the solvent-accessible surface area (FISA) describe the IC(50) of 19 inhibitors of the PC-3 cell line with activities ranging over five orders of magnitude with an R(2)=0.92 and F=59.2. This information will be helpful in the further design of novel anticancer drugs for treatment of prostate cancer and other diseases affected by HDAC inhibition.

Analyses of Activity for Factor Xa Inhibitors Based on Monte Carlo Simulations

Monte Carlo/Extended Linear Response (MC/ELR) simulations have been conducted on 60 inhibitors of human factor Xa to determine the important interactions associated with their activity. A variety of physicochemical descriptors were configurationally averaged during the course of the simulations of each inhibitor bound to factor Xa and free in water. A regression equation was then derived; it reproduces the experimental inhibition data with a correlation coefficient, r(2), of 0.74, an rms error of 0.67 kcal/mol, and an average unsigned error of 0.60 kcal/mol using only two physically reasonable descriptors. The two factors that emerged as important in determining inhibitory potential are (1) favorable van der Waals interactions between protein and ligand and (2) direct hydrogen bonding between the inhibitor and protein. The conclusions were supported with structural analyses and results of MC/free energy perturbation (FEP) calculations.

A General Treatment of Solubility. 1. The QSPR Correlation of Solvation Free Energies of Single Solutes in Series of Solvents

We present an extended QSPR modeling of solubilities of about 500 substances in series of up to 69 diverse solvents. The models are obtained with our new software package, CODESSA PRO, which is furnished with an advanced variable selection procedure and a large pool of theoretically derived molecular descriptors. The squared correlation coefficients and squared standard deviations (variances) range from 0.837 and 0.1 for 2-pyrrolidone to 0.998 and 0.02 for dipropyl ether, respectively. The predictive power of the models was verified by using the "leave-one-out" cross-validation procedure. The QSPR models presented are suitable for the rapid evaluation of solvation free energies of organic compounds.

GSSI, a General Model for Solute−Solvent Interactions. 1. Description of the Model

A novel, semiempirical approach for the general treatment of solute-solvent interactions (GSSI) was developed to enable the prediction of solution-phase properties (e.g., free energies of desolvation, partition coefficients, and membrane permeabilities). The GSSI approach is based on the principle that all solution-phase processes can be modeled in terms of one or more gas-to-solution transfer processes. The free energy of each gas-to-solution transfer process is calculated as the sum of the free energy of cavity formation and the free energy of solute-solvent interaction. The solute's contributions to these free energies are modeled on the basis of various quantities computed from the solute's three-dimensional (3D) structure, whereas the solvent's contributions are modeled by empirically determined regression coefficients. More specifically, the free energy of cavity formation is modeled on the basis of the total solvent-accessible surface area of the solute. The enthalpy of solute-solvent interaction is modeled on the basis of intermolecular interaction potentials calculated at many uniformly distributed points on the solvent-accessible surface of the solute. The entropy of solute-solvent interaction is modeled on the basis of an effective number of rotatable bonds in the solute and by the regression coefficients characteristic of the solvent. The potential utility of the GSSI approach was demonstrated by modeling the free energy of gas-to-solution transfer for 111 solutes in water, 250 solutes in hexadecane, and 84 solutes in octanol.

Continuum and discrete calculation of fractional contributions to solvation free energy

Approaches to compute fractional contributions to the solvation free energy are developed in the context of continuum self consistent reaction field calculations (both classical and quantum mechanical), as well as in the framework of discrete molecular dynamics simulations. It is found that for a series of typical pharmacological drugs there is a good agreement between the different fractional descriptions. Algorithms reported here can be easily applied as molecular descriptors in the context of quantitative structure-activity relationships.

Molecular modeling calculations of HIV-1 reverse transcriptase nonnucleoside inhibitors: correlation of binding energy with biological activity for novel 2-aryl-substituted benzimidazole analogues

The energies and physical descriptors for the binding of 20 novel 1-(2,6-difluorobenzyl)-2-(2,6-difluorophenyl)benzimidazole analogues (BPBIs) to HIV-1 reverse transcriptase (RT) have been determined using Monte Carlo (MC) simulations. The crystallographic structure of the lead compound, 1-(2,6-difluorobenzyl)-2-(2,6-difluorophenyl)-4-methylbenzimidazole, was used as a starting point to model the inhibitors in both the bound and the unbound states. The energy terms and physical descriptors obtained from the calculations were correlated with their respective experimental EC(50) values, resulting in an r(2) value of 0.70 and a root-mean-square deviation (rms) of 0.53 kcal/mol. The terms in the correlation include the change in total Coulombic energy and solvent-accessible surface area. Structural analysis of the data files from the BPBI calculations reveals that all of the analogues with good biological activity show the formation of a hydrogen bond between the ligand and the backbone nitrogen atom of lysine 103. By use of the structural results, two novel BPBI inhibitors have been designed and calculations have been carried out. The results show the formation of the desired hydrogen bonds, and the DeltaG(binding) values predict the compounds to be excellent RT inhibitors. Subsequent synthesis and biological activity testing of these analogues have shown the validity of the predictive calculations. If the BPBIs are modeled in a site constructed from the crystal coordinates of a member of another class of nonnucleoside inhibitors (the 4,5,6,7-tetrahydroimidazo[4,5,1-jk][1,4]benzodiazepine-2(1H)-thione and -one (TIBO) compounds), the correlation with the same terms drops slightly, giving an r(2) value of 0.61 with an associated root-mean-square value of 0.53 kcal/mol. Conversely, if the TIBO compounds are modeled in a site constructed from the BPBI complex crystal coordinates, a correlation can be obtained using the drug-protein interaction energy and change in the total number of hydrogen bonds, giving an r(2) value of 0.63. These are the same descriptors that were used for the TIBO compounds modeled in their own sites, where the r(2) value was 0.72. These data suggest that it may be possible, in some cases, to design novel inhibitors utilizing structural data from related, but not identical, inhibitors.

Transferability of fragmental contributions to the octanol/water partition coefficient: an NDDO-based MST study

This study examines the transferability of fragmental contributions to the octanol/water partition coefficient. As a previous step, we report the parameterization of the AM1 and PM3 versions of the MST model for n-octanol. The final AM1 and PM3 MST models reproduce the experimental free energy of solvation and the octanol/water partition coefficient (log P(ow)) with a root-mean-square deviation of around 0.7 kcal/mol and 0.5 (in units of log P), respectively. Based on this parameterization, an NNDO-based procedure is presented to dissect the free energy of transfer between octanol and water in contributions directly associated with specific atoms or functional groups. The application of this procedure to a set of representative molecular systems illustrates the dependence of the log P(ow) fragmental contribution due to electronic, hydrogen bonding, and steric effects, which cannot be easily accounted for in simple additive-based empirical schemes. The results point out the potential use of theoretical methods to refine the fragmental contributions in empirical methods.

Fast, efficient generation of high-quality atomic charges. AM1-BCC model: II. Parameterization and validation

We present the first global parameterization and validation of a novel charge model, called AM1-BCC, which quickly and efficiently generates high-quality atomic charges for computer simulations of organic molecules in polar media. The goal of the charge model is to produce atomic charges that emulate the HF/6-31G* electrostatic potential (ESP) of a molecule. Underlying electronic structure features, including formal charge and electron delocalization, are first captured by AM1 population charges; simple additive bond charge corrections (BCCs) are then applied to these AM1 atomic charges to produce the AM1-BCC charges. The parameterization of BCCs was carried out by fitting to the HF/6-31G* ESP of a training set of >2700 molecules. Most organic functional groups and their combinations were sampled, as well as an extensive variety of cyclic and fused bicyclic heteroaryl systems. The resulting BCC parameters allow the AM1-BCC charging scheme to handle virtually all types of organic compounds listed in The Merck Index and the NCI Database. Validation of the model was done through comparisons of hydrogen-bonded dimer energies and relative free energies of solvation using AM1-BCC charges in conjunction with the 1994 Cornell et al. forcefield for AMBER.(13) Homo- and hetero-dimer hydrogen-bond energies of a diverse set of organic molecules were reproduced to within 0.95 kcal/mol RMS deviation from the ab initio values, and for DNA dimers the energies were within 0.9 kcal/mol RMS deviation from ab initio values. The calculated relative free energies of solvation for a diverse set of monofunctional isosteres were reproduced to within 0.69 kcal/mol of experiment. In all these validation tests, AMBER with the AM1-BCC charge model maintained a correlation coefficient above 0.96. Thus, the parameters presented here for use with the AM1-BCC method present a fast, accurate, and robust alternative to HF/6-31G* ESP-fit charges for general use with the AMBER force field in computer simulations involving organic small molecules.

Prediction of activity for nonnucleoside inhibitors with HIV-1 reverse transcriptase based on Monte Carlo simulations

Results of Monte Carlo (MC) simulations for more than 200 nonnucleoside inhibitors of HIV-1 reverse transcriptase (NNRTIs) representing eight diverse chemotypes have been correlated with their anti-HIV activities in an effort to establish simulation protocols and methods that can be used in the development of more effective drugs. Each inhibitor was modeled in a complex with the protein and by itself in water, and potentially useful descriptors of binding affinity were collected during the MC simulations. A viable regression equation was obtained for each data set using an extended linear response approach, which yielded r(2) values between 0.54 and 0.85 and an average unsigned error of only 0.50 kcal/mol. The most common descriptors confirm that a good geometrical match between the inhibitor and the protein is important and that the net loss of hydrogen bonds with the inhibitor upon binding is unfavorable. Other physically reasonable descriptors of binding are needed on a chemotype case-by-case basis. By including descriptors in common from the individual fits, combination regressions that include multiple data sets were also developed. This procedure led to a refined "master" regression for 210 NNRTIs with an r(2) of 0.60 and a cross-validated q(2) of 0.55. The computed activities show an rms error of 0.86 kcal/mol in comparison with experiment and an average unsigned error of 0.69 kcal/mol. Encouraging results were obtained for the predictions of 27 NNRTIs, representing a new chemotype not included in the development of the regression model. Predictions for this test set using the master regression yielded a q(2) value of 0.51 and an average unsigned error of 0.67 kcal/mol. Finally, additional regression analysis reveals that use of ligand-only descriptors leads to models with much diminished predictive ability.

Structural requirements for inhibition of the neuronal nitric oxide synthase (NOS-I): 3D-QSAR analysis of 4-oxo- and 4-amino-pteridine-based inhibitors

The family of homodimeric nitric oxide synthases (NOS I-III) catalyzes the generation of the cellular messenger nitric oxide (NO) by oxidation of the substrate L-arginine. The rational design of specific NOS inhibitors is of therapeutic interest in regulating pathological NO levels associated with sepsis, inflammatory, and neurodegenerative diseases. The cofactor (6R)-5,6,7,8-tetrahydrobiopterin (H(4)Bip) maximally activates all NOSs and stabilizes enzyme quaternary structure by promoting and stabilizing dimerization. Here, we describe the synthesis and three-dimensional (3D) quantitative structure-activity relationship (QSAR) analysis of 65 novel 4-amino- and 4-oxo-pteridines (antipterins) as inhibitors targeting the H(4)Bip binding site of the neuronal NOS isoform (NOS-I). The experimental binding modes for two inhibitors complexed with the related endothelial NO synthase (NOS-III) reveal requirements of biological affinity and form the basis for ligand alignment. Different alignment rules were derived by building other compounds accordingly using manual superposition or a genetic algorithm for flexible superposition. Those alignments led to 3D-QSAR models (comparative molecular field analysis (CoMFA) and comparative molecular similarity index analysis (CoMSIA)), which were validated using leave-one-out cross-validation, multiple analyses with two and five randomly chosen cross-validation groups, perturbation of biological activities by randomization or progressive scrambling, and external prediction. An iterative realignment procedure based on rigid field fit was used to improve the consistency of the resulting partial least squares models. This led to consistent and highly predictive 3D-QSAR models with good correlation coefficients for both CoMFA and CoMSIA, which correspond to experimentally determined NOS-II and -III H(4)Bip binding site topologies as well as to the NOS-I homology model binding site in terms of steric, electrostatic, and hydrophobic complementarity. These models provide clear guidelines and accurate activity predictions for novel NOS-I inhibitors.

Fast prediction of hydration free energies from molecular interaction fields

A novel empirical model is presented that allows the fast computation of hydration free energies with high accuracy. The linear model is based upon the separation of the free energy of hydration into a cavity and an interaction term. The cavity contribution is modeled as a linear combination of molecular volume and surface terms. The interaction part is derived from the statistical three-dimensional (3D) free energy density and is modeled approximately as a molecular interaction field using the program GRID. A compression scheme is employed to represent this 3D information on the molecular surface by means of a linear combination of surface functions. A set of 81 small organic molecules with known experimental hydration free energies is used to determine the coefficients of the linear model by least squares regression. The fit is statistically significant yielding a correlation coefficient of 0.99, a root mean square error of 0.27 kcal/mol for the 81 molecules belonging to the training set, and 0.63 kcal/mol for an independent test set of 10 molecules.

Prediction of drug solubility from structure

The aqueous solubility of a drug is an important factor affecting its bioavailability. Numerous computational methods have been developed for the prediction of aqueous solubility from a compound's structure. A review is provided of the methodology and quality of results for the most useful procedures including the model implemented in the QikProp program. Viable methods now exist for predictions with less than 1 log unit uncertainty, which is adequate for prescreening synthetic candidates or design of combinatorial libraries. Further progress with predictive methods would require an experimental database of highly accurate solubilities for a large, diverse collection of drug-like molecules.

Prediction of aqueous solubility of drugs and pesticides with COSMO-RS

The COSMO-RS method, originally developed for the prediction of liquid-liquid and liquid-vapor equilibrium constants based on quantum chemical calculations, has been extended to solid compounds by addition of a heuristic expression for the Gibbs free energy of fusion. By this addition, COSMO-RS is now capable of a priori prediction of aqueous solubilities of a wide range of typical neutral drug and pesticide compounds. Only three parameters in the heuristic expression have been fitted on a data set of 150 drug-like compounds. On these data an rms deviation of 0.66 log-units was achieved. Later, the model was tested on a set of 107 pesticides, which have been critically selected based on two experimental data sources and by a crosscheck with an independent HQSAR model. On this data set an rms of 0.61 log-units was achieved, without any adjustments to the structurally extremely diverse pesticides. This result verifies the ability of this extended COSMO-RS to predict aqueous solubilities of drugs and pesticides of almost arbitrary structural classes. The new method is COSMO-RSol.

Rational determination of transfer free energies of small drugs across the water-oil interface

The application of statistical simulations to the estimation of transfer free energies of pharmacologically relevant organic molecules is reported. Large-scale molecular dynamics simulations have been carried out on a series of four solutes, viz. antipyrine, caffeine, ganciclovir, and alpha-D-glucose, at the water-dodecane interface as a model of a biological water-membrane interfacial system. Agreement with experimentally determined partition coefficients is remarkable, demonstrating that free energy calculations, when executed with appropriate protocols, have reached the maturity to predict thermodynamic quantities of interest to the pharmaceutical world. The computational effort that warrants accurate, converged free energies remains, however, in large measure, incompatible with the high-throughput exploration of large sets of pharmacologically active drugs sought by industrial settings. Compared to the cost-effective, fast estimation of simple partition coefficients, the present free energy calculations, nevertheless, offer a far more detailed information about the underlying energetics of the system when the solute is translocated across the water-dodecane interface, which can be valuable in the context of de novo drug design.

The impact of informatics and computational chemistry on synthesis and screening

High-throughput synthesis and screening technologies have enhanced the impact of computational chemistry on the drug discovery process. From the design of targeted, drug-like libraries to 'virtual' optimization of potency, selectivity and ADME/Tox properties, computational chemists are able to efficiently manage costly resources and dramatically shorten drug discovery cycle times. This review will describe some of the successful strategies and applications of state-of-the-art algorithms to enhance drug discovery, as well as key points in the drug discovery process where computational methods can have, and have had, greatest impact.

Influence of highly polyunsaturated lipid acyl chains of biomembranes on the NMR order parameters

We investigate the effect of specific conformations of double-bond segments in highly polyunsaturated acyl chains on the deuterium (2)H NMR order parameters of a fully hydrated 1-stearoyl-2-docosahexaenoyl-sn-glycero-3-phosphocholine (SDPC, 18:0/22:6 PC) lipid bilayer. The system is analyzed by performing a molecular dynamics simulation study at ambient conditions in the fluid lamellar phase. By separately calculating the different partial contributions to the total order parameter profiles measurable experimentally, we are able to get insights into the molecular origin of earlier experimental and theoretical observations. The effect of the position of the different conformations of double-bond segments along the polyunsaturated acyl chain is also examined. As in experiments performed in a series of lipid bilayers with an increasing number of cis double bonds per lipid molecule [Holte, L. L., et al. Biophys. J. 1995, 68, 2396], we find that unsaturations influence mainly the order of the bottom half of the saturated chain. Specific conformations of the polyunsaturated chain close to the lipid headgroups have a distinct effect on the order of the bottom half of the saturated chain and on the top half of the polyunsaturated chain. Our results indicate that for SDPC the conformation of the region of the polyunsaturated chain located between the first three cis double bonds is responsible for the major effects on the orientational order of both the saturated and the polyunsaturated chains.

Estimation of binding affinities for selective thrombin inhibitors via Monte Carlo simulations

Monte Carlo simulations have been performed on a series of 20 active-site-directed thrombin inhibitors to determine the interactions and energetics associated with the binding of these compounds. Physicochemical descriptors of potential value in the prediction of binding affinities were averaged during simulations of each inhibitor unbound in water and bound to thrombin. Regression equations based on 3-5 descriptors are able to reproduce the experimental binding affinities, which cover a 7 kcal/mol range, with rms errors of 1.0-1.3 kcal/mol, and yield correlation coefficients, r(2), of 0.7-0.8. On the basis of these results, the quantities most important in determining the binding affinities are: (1) the enhancement of van der Waals interactions in going from solution to the bound state, (2) the intramolecular strain induced in the inhibitor upon binding, (3) the number of hydrogen bonds lost in the binding process, and (4) the number of rotatable bonds in the inhibitor. The descriptors are physically reasonable and, in combination with the insights gained from analysis of the simulation structures, suggest directions for the development of improved thrombin inhibitors.

Estimation of binding affinities for HEPT and nevirapine analogues with HIV-1 reverse transcriptase via Monte Carlo simulations

The interactions and energetics associated with the binding of 20 HEPT and 20 nevirapine nonnucleoside inhibitors of HIV-1 reverse transcriptase (RT) have been explored in an effort to establish simulation protocols and methods that can be used in the development of more effective anti-HIV drugs. Using crystallographic structures as starting points, all 40 inhibitors were modeled in the bound and unbound states via Monte Carlo (MC) statistical mechanics methods. Potentially useful descriptors of binding affinity were configurationally averaged for each inhibitor during the MC simulations, and correlations were sought with reported experimental activities. A viable regression equation was obtained using only four descriptors to correlate the 40 experimental activities with an r(2)() of 0.75 and cross-validated q(2)() of 0.69. The computed activities show a rmsd of 0.94 kcal/mol in comparison with experiment and an average unsigned error of 0.69 kcal/mol. The MC results reveal three physically reasonable parameters that control the binding affinities: (1) loss of hydrogen bonds with the inhibitor is unfavorable, (2) burial of hydrophobic surface area is favorable, and (3) a good geometrical fit without steric clashes is needed for the protein-inhibitor complex. It is gratifying that the corresponding descriptors are statistically the most important quantities for determining the anti-HIVRT activity for the 40 compounds. Representative examples are also given in which structural and thermodynamic information from the MC simulations is used to help understand binding differences for related compounds. A key pi-type hydrogen bond has been identified between secondary-amide nevirapine analogues and Tyr188A of HIVRT that explains their otherwise surprising activity and the ineffectiveness of nevirapine against the Y188C mutant.