Further development and validation of empirical scoring functions for structure-based binding affinity prediction

New empirical scoring functions have been developed to estimate the binding affinity of a given protein-ligand complex with known three-dimensional structure. These scoring functions include terms accounting for van der Waals interaction, hydrogen bonding, deformation penalty, and hydrophobic effect. A special feature is that three different algorithms have been implemented to calculate the hydrophobic effect term, which results in three parallel scoring functions. All three scoring functions are calibrated through multivariate regression analysis of a set of 200 protein-ligand complexes and they reproduce the binding free energies of the entire training set with standard deviations of 2.2 kcal/mol, 2.1 kcal/mol, and 2.0 kcal/mol, respectively. These three scoring functions are further combined into a consensus scoring function, X-CSCORE. When tested on an independent set of 30 protein-ligand complexes, X-CSCORE is able to predict their binding free energies with a standard deviation of 2.2 kcal/mol. The potential application of X-CSCORE to molecular docking is also investigated. Our results show that this consensus scoring function improves the docking accuracy considerably when compared to the conventional force field computation used for molecular docking.     

Robust scoring functions for protein-ligand interactions with quantum chemical charge models

Ordinary least-squares (OLS) regression has been used widely for constructing the scoring functions for protein-ligand interactions. However, OLS is very sensitive to the existence of outliers, and models constructed using it are easily affected by the outliers or even the choice of the data set. On the other hand, determination of atomic charges is regarded as of central importance, because the electrostatic interaction is known to be a key contributing factor for biomolecular association. In the development of the AutoDock4 scoring function, only OLS was conducted, and the simple Gasteiger method was adopted. It is therefore of considerable interest to see whether more rigorous charge models could improve the statistical performance of the AutoDock4 scoring function. In this study, we have employed two well-established quantum chemical approaches, namely the restrained electrostatic potential (RESP) and the Austin-model 1-bond charge correction (AM1-BCC) methods, to obtain atomic partial charges, and we have compared how different charge models affect the performance of AutoDock4 scoring functions. In combination with robust regression analysis and outlier exclusion, our new protein-ligand free energy regression model with AM1-BCC charges for ligands and Amber99SB charges for proteins achieve lowest root-mean-squared error of 1.637 kcal/mol for the training set of 147 complexes and 2.176 kcal/mol for the external test set of 1427 complexes. The assessment for binding pose prediction with the 100 external decoy sets indicates very high success rate of 87% with the criteria of predicted root-mean-squared deviation of less than 2 ?. The success rates and statistical performance of our robust scoring functions are only weakly class-dependent (hydrophobic, hydrophilic, or mixed).     

SLICK--scoring and energy functions for protein-carbohydrate interactions

Protein-carbohydrate interactions are increasingly being recognized as essential for many important biomolecular recognition processes. From these, numerous biomedical applications arise in areas as diverse as drug design, immunology, or drug transport. We introduce SLICK, a package containing a scoring and an energy function, which were specifically designed to predict binding modes and free energies of sugars and sugarlike compounds to proteins. SLICK accounts for van der Waals interactions, solvation effects, electrostatics, hydrogen bonds, and CH...pi interactions, the latter being a particular feature of most protein-carbohydrate interactions. Parameters for the empirical energy function were calibrated on a set of high-resolution crystal structures of protein-sugar complexes with known experimental binding free energies. We show that SLICK predicts the binding free energies of predicted complexes (through molecular docking) with high accuracy. SLICK is available as part of our molecular modeling package BALL (www.ball-project.org).     

Universal solvation model based on conductor‐like screening model

Atomic surface tensions are parameterized for use with solvation models in which the electrostatic part of the calculation is based on the conductor‐like screening model (COSMO) and the semiempirical molecular orbital methods AM1, PM3, and MNDO/d. The convergence of the calculated polarization free energies with respect to the numerical parameters of the electrostatic calculations is first examined. The accuracy and precision of the calculated values are improved significantly by adjusting two parameters that control the segmentation of the solvent‐accessible surface that is used for the calculations. The accuracy of COSMO calculations is further improved by adopting an optimized set of empirical electrostatic atomic radii. Finally, the electrostatic calculation is combined with SM5‐type atomic surface tension functionals that are used to compute the nonelectrostatic portions of the solvation free energy. All parameterizations are carried out using rigid (R) gas‐phase geometries; this combination (SM5‐type surface tensions, COSMO electrostatics, and rigid geometries) is called SM5CR. Six air–water and 76 water–solvent partition coefficients are added to the training set of air–solvent data points previously used to parameterize the SM5 suite of solvation models, thereby bringing the total number of data points in the training set to 2266. The model yields free energies of solvation and transfer with mean unsigned errors of 0.63, 0.59, and 0.61 kcal/mol for AM1, PM3, and MNDO/d, respectively, over all 2217 data points for neutral solutes in the training set and mean unsigned errors of 3.0, 2.7, and 3.1 kcal/mol, respectively, for 49 data points for the ions. © 2000 John Wiley & Sons, Inc. J Comput Chem 21: 340–366, 2000     

Solvated interaction energy (SIE) for scoring protein-ligand binding affinities. 2. Benchmark in the CSAR-2010 scoring exercise

Solvated interaction energy (SIE) is an end-point physics-based scoring function for predicting binding affinities from force-field nonbonded interaction terms, continuum solvation, and configurational entropy linear compensation. We tested the SIE function in the Community Structure-Activity Resource (CSAR) scoring challenge consisting of high-resolution cocrystal structures for 343 protein-ligand complexes with high-quality binding affinity data and high diversity with respect to protein targets. Particular emphasis was placed on the sensitivity of SIE predictions to the assignment of protonation and tautomeric states in the complex and the treatment of metal ions near the protein-ligand interface. These were manually curated from an originally distributed CSAR-HiQ data set version, leading to the currently distributed CSAR-NRC-HiQ version. We found that this manual curation was a critical step for accurately testing the performance of the SIE function. The standard SIE parametrization, previously calibrated on an independent data set, predicted absolute binding affinities with a mean-unsigned-error (MUE) of 2.41 kcal/mol for the CSAR-HiQ version, which improved to 1.98 kcal/mol for the upgraded CSAR-NRC-HiQ version. Half-half retraining-testing of SIE parameters on two predefined subsets of CSAR-NRC-HiQ led to only marginal further improvements to an MUE of 1.83 kcal/mol. Hence, we do not recommend altering the current default parameters of SIE at this time. For a sample of SIE outliers, additional calculations by molecular dynamics-based SIE averaging with or without incorporation of ligand strain, by MM-PB(GB)/SA methods with or without entropic estimates, or even by the linear interaction energy (LIE) formalism with an explicit solvent model, did not further improve predictions.     

Continuum polarizable force field within the Poisson-Boltzmann framework

We have developed and tested a complete set of nonbonded parameters for a continuum polarizable force field. Our analysis shows that the new continuum polarizable model is consistent with B3LYP/cc-pVTZ in modeling electronic response upon variation of dielectric environment. Comparison with experiment also shows that the new continuum polarizable model is reasonable, with accuracy similar to that of B3LYP/cc-pVTZ in reproduction of dipole moments of selected organic molecules in the gas phase. We have further tested the validity to interchange the Amber van der Waals parameters between the explicit and continuum polarizable force fields with a series of dimers. It can be found that the continuum polarizable model agrees well with MP2/cc-pVTZ, with deviations in dimer binding energies less than 0.9 kcal/mol in the aqueous dielectric environment. Finally, we have optimized atomic cavity radii with respect to experimental solvation free energies of 177 training molecules. To validate the optimized cavity radii, we have tested these parameters against 176 test molecules. It is found that the optimized Poisson-Boltzmann atomic cavity radii transfer well from the training set to the test set, with an overall root-mean-square deviation of 1.30 kcal/mol, an unsigned average error of 1.07 kcal/mol, and a correlation coefficient of 92% for all 353 molecules in both the training and test sets. Given the development documented here, the next natural step is the construction of a full protein/nucleic acid force field within the new continuum polarization framework.     

Calculation of Host-Guest Binding Affinities Using a Quantum-Mechanical Energy Model

The prediction of protein-ligand binding affinities is of central interest in computer-aided drug discovery, but it is still difficult to achieve a high degree of accuracy. Recent studies suggesting that available force fields may be a key source of error motivate the present study, which reports the first mining minima (M2) binding affinity calculations based on a quantum mechanical energy model, rather than an empirical force field. We apply a semi-empirical quantum-mechanical energy function, PM6-DH+, coupled with the COSMO solvation model, to 29 host-guest systems with a wide range of measured binding affinities. After correction for a systematic error, which appears to derive from the treatment of polar solvation, the computed absolute binding affinities agree well with experimental measurements, with a mean error 1.6 kcal/mol and a correlation coefficient of 0.91. These calculations also delineate the contributions of various energy components, including solute energy, configurational entropy, and solvation free energy, to the binding free energies of these host-guest complexes. Comparison with our previous calculations, which used empirical force fields, point to significant differences in both the energetic and entropic components of the binding free energy. The present study demonstrates successful combination of a quantum mechanical Hamiltonian with the M2 affinity method.     

Extension of the PDDG/PM3 and PDDG/MNDO semiempirical molecular orbital methods to the halogens

The new semiempirical methods, PDDG/PM3 and PDDG/MNDO, have been parameterized for halogens. For comparison, the original MNDO and PM3 were also reoptimized for the halogens using the same training set; these modified methods are referred to as MNDO' and PM3'. For 442 halogen-containing molecules, the smallest mean absolute error (MAE) in heats of formation is obtained with PDDG/PM3 (5.6 kcal/mol), followed by PM3' (6.1 kcal/mol), PDDG/MNDO (6.6 kcal/mol), PM3 (8.1 kcal/mol), MNDO' (8.5 kcal/mol), AM1 (11.1 kcal/mol), and MNDO (14.0 kcal/mol). For normal-valent halogen-containing molecules, the PDDG methods also provide improved heats of formation over MNDO/d. Hypervalent compounds were not included in the training set and improvements over the standard NDDO methods with sp basis sets were not obtained. For small haloalkanes, the PDDG methods yield more accurate heats of formation than are obtained from density functional theory (DFT) with the B3LYP and B3PW91 functionals using large basis sets. PDDG/PM3 and PM3' also give improved binding energies over the standard NDDO methods for complexes involving halide anions, and they are competitive with B3LYP/6-311++G(d,p) results including thermal corrections. Among the semiempirical methods studied, PDDG/PM3 also generates the best agreement with high-level ab initio G2 and CCSD(T) intrinsic activation energies for S(N)2 reactions involving methyl halides and halide anions. Finally, the MAEs in ionization potentials, dipole moments, and molecular geometries show that the parameter sets for the PDDG and reoptimized NDDO methods reduce the MAEs in heats of formation without compromising the other important QM observables.     

PLASS: Protein-ligand affinity statistical score – a knowledge-based force-field model of interaction derived from the PDB

We have developed PLASS (Protein-Ligand Affinity Statistical Score), a pair-wise potential of mean-force for rapid estimation of the binding affinity of a ligand molecule to a protein active site. This scoring function is derived from the frequency of occurrence of atom-type pairs in crystallographic complexes taken from the Protein Data Bank (PDB). Statistical distributions are converted into distance-dependent contributions to the Gibbs free interaction energy for 10 atomic types using the Boltzmann hypothesis, with only one adjustable parameter. For a representative set of 72 protein-ligand structures, PLASS scores correlate well with the experimentally measured dissociation constants: a correlation coefficient R of 0.82 and RMS error of 2.0 kcal/mol. Such high accuracy results from our novel treatment of the volume correction term, which takes into account the inhomogeneous properties of the protein-ligand complexes. PLASS is able to rank reliably the affinity of complexes which have as much diversity as in the PDB.     

Predictions of hydration free energies from continuum solvent with solute polarizable models: the SAMPL2 blind challenge

This paper reports the results of our attempt to predict hydration free energies on the SAMPL2 blind challenge dataset. We mostly examine the effects of the solute electrostatic component on the accuracy of the predictions. The usefulness of electronic polarization in predicting hydration free energies is assessed by comparing the Electronic Polarization from Internal Continuum model and the self consistent reaction field IEF-PCM to standard non-polarizable charge models such as RESP and AM1-BCC. We also determine an optimal restraint weight for Dielectric-RESP atomic charges fitting. Statistical analysis of the results could not distinguish the methods from one another. The smallest average unsigned error obtained is 1.9 ± 0.6 kcal/mol (95% confidence level). A class of outliers led us to investigate the importance of the solute–solvent instantaneous induction energy, a missing term in PB continuum models. We estimated values between −1.5 and −6 kcal/mol for a series of halo-benzenes which can explain why some predicted hydration energies of non-polar molecules significantly disagreed with experiment.     

A Mixed QM/MM Scoring Function to Predict Protein-Ligand Binding Affinity

Computational methods for predicting protein-ligand binding free energy continue to be popular as a potential cost-cutting method in the drug discovery process. However, accurate predictions are often difficult to make as estimates must be made for certain electronic and entropic terms in conventional force field based scoring functions. Mixed quantum mechanics/molecular mechanics (QM/MM) methods allow electronic effects for a small region of the protein to be calculated, treating the remaining atoms as a fixed charge background for the active site. Such a semi-empirical QM/MM scoring function has been implemented in AMBER using DivCon and tested on a set of 23 metalloprotein-ligand complexes, where QM/MM methods provide a particular advantage in the modeling of the metal ion. The binding affinity of this set of proteins can be calculated with an R(2) of 0.64 and a standard deviation of 1.88 kcal/mol without fitting and 0.71 and a standard deviation of 1.69 kcal/mol with fitted weighting of the individual scoring terms. In this study we explore using various methods to calculate terms in the binding free energy equation, including entropy estimates and minimization standards. From these studies we found that using the rotational bond estimate to ligand entropy results in a reasonable R(2) of 0.63 without fitting. We also found that using the ESCF energy of the proteins without minimization resulted in an R(2) of 0.57, when using the rotatable bond entropy estimate.     

Novel, customizable scoring functions, parameterized using N-PLS, for structure-based drug discovery

The ability to accurately predict biological affinity on the basis of in silico docking to a protein target remains a challenging goal in the CADD arena. Typically, "standard" scoring functions have been employed that use the calculated docking result and a set of empirical parameters to calculate a predicted binding affinity. To improve on this, we are exploring novel strategies for rapidly developing and tuning "customized" scoring functions tailored to a specific need. In the present work, three such customized scoring functions were developed using a set of 129 high-resolution protein-ligand crystal structures with measured Ki values. The functions were parametrized using N-PLS (N-way partial least squares), a multivariate technique well-known in the 3D quantitative structure-activity relationship field. A modest correlation between observed and calculated pKi values using a standard scoring function (r2 = 0.5) could be improved to 0.8 when a customized scoring function was applied. To mimic a more realistic scenario, a second scoring function was developed, not based on crystal structures but exclusively on several binding poses generated with the Flo+ docking program. Finally, a validation study was conducted by generating a third scoring function with 99 randomly selected complexes from the 129 as a training set and predicting pKi values for a test set that comprised the remaining 30 complexes. Training and test set r2 values were 0.77 and 0.78, respectively. These results indicate that, even without direct structural information, predictive customized scoring functions can be developed using N-PLS, and this approach holds significant potential as a general procedure for predicting binding affinity on the basis of in silico docking.     

Improved interaction potentials for charged residues in proteins

Electrostatic interactions dominate the structure and free energy of biomolecules. To obtain accurate free energies involving charged groups from molecular simulations, OPLS-AA parameters have been reoptimized using Monte Carlo free energy perturbation. New parameters fit a self-consistent, experimental set of hydration free energies for acetate (Asp), propionate (Glu), 4-methylimidazolium (Hip), n-butylammonium (Lys), and n-propylguanidinium (Arg), all resembling charged residue side chains, including beta-carbons. It is shown that OPLS-AA free energies depend critically on the type of water model, TIP4P or TIP3P; i.e., each water model requires specific water-charged molecule interaction potentials. New models (models 1 and 3) are thus described for both water models. Uncertainties in relative free energies of charged residues are approximately 2 kcal/mol with the new parameters, due to variations in system setup (MAEs of ca. 1 kcal/mol) and noise from simulations (ca. 1 kcal/mol). The latter error of approximately 1 kcal/mol contrasts MAEs from standard OPLS-AA of up to 13 kcal/mol for the entire series of charged residues or up to 5 kcal/mol for the cationic series Lys, Arg, and Hip. The new parameters can be used directly in molecular simulations with no modification of neutral residues needed and are envisioned to be particular important in simulations where charged residues change environment.     

Solvated interaction energy (SIE) for scoring protein-ligand binding affinities. 1. Exploring the parameter space

We present a binding free energy function that consists of force field terms supplemented by solvation terms. We used this function to calibrate the solvation model along with the binding interaction terms in a self-consistent manner. The motivation for this approach was that the solute dielectric-constant dependence of calculated hydration gas-to-water transfer free energies is markedly different from that of binding free energies (J. Comput. Chem. 2003, 24, 954). Hence, we sought to calibrate directly the solvation terms in the context of a binding calculation. The five parameters of the model were systematically scanned to best reproduce the absolute binding free energies for a set of 99 protein-ligand complexes. We obtained a mean unsigned error of 1.29 kcal/mol for the predicted absolute binding affinity in a parameter space that was fairly shallow near the optimum. The lowest errors were obtained with solute dielectric values of Din = 20 or higher and scaling of the intermolecular van der Waals interaction energy by factors ranging from 0.03 to 0.15. The high apparent Din and strong van der Waals scaling may reflect the anticorrelation of the change in solvated potential energy and configurational entropy, that is, enthalpy-entropy compensation in ligand binding (Biophys. J. 2004, 87, 3035-3049). Five variations of preparing the protein-ligand data set were explored in order to examine the effect of energy refinement and the presence of bound water on the calculated results. We find that retaining water in the final protein structure used for calculating the binding free energy is not necessary to obtain good results; that is the continuum solvation model is sufficient. Virtual screening enrichment studies on estrogen receptor and thymidine kinase showed a good ability of the binding free energy function to recover true hits in a collection of decoys.     

Predicting binding affinities of host-guest systems in the SAMPL3 blind challenge: the performance of relative free energy calculations

Relative free energy calculations based on molecular dynamics simulations are combined with available experimental binding free energies to predict unknown binding affinities of acyclic Cucurbituril complexes in the blind SAMPL3 competition. The predictions yield root mean square errors between 2.6 and 3.2 kcal/mol for seven host-guest systems. Those deviations are comparable to results for solvation free energies of small organic molecules. However, the standard deviations found in our simulations range from 0.4 to 2.4 kcal/mol, which indicates the need for better sampling. Three different approaches are compared. Bennett's Acceptance Ratio Method and thermodynamic integration based on the trapezoidal rule with 12 λ-points exhibit a root mean square error of 2.6 kcal/mol, while thermodynamic integration with Simpson's rule and 11 λ-points leads to a root mean square error of 3.2 kcal/mol. In terms of absolute median errors, Bennett's Acceptance Ratio Method performs better than thermodynamic integration with the trapezoidal rule (1.7 vs. 2.9 kcal/mol). Simulations of the deprotonated forms of the guest molecules exhibit a poorer correspondence to experimental results with a root mean square error of 5.2 kcal/mol. In addition, a decrease of the buffer concentration by approximately 20 mM in the simulations raises the root mean square error to 3.8 kcal/mol.     

Empirical scoring functions: I. The development of a fast empirical scoring function to estimate the binding affinity of ligands in receptor complexes

This paper describes the development of a simple empirical scoringfunction designed to estimate the free energy of binding for aprotein–ligand complex when the 3D structure of the complex is knownor can be approximated. The function uses simple contact terms to estimatelipophilic and metal–ligand binding contributions, a simple explicitform for hydrogen bonds and a term which penalises flexibility. Thecoefficients of each term are obtained using a regression based on 82ligand–receptor complexes for which the binding affinity is known. Thefunction reproduces the binding affinity of the complexes with across-validated error of 8.68 kJ/mol. Tests on internal consistency indicatethat the coefficients obtained are stable to changes in the composition ofthe training set. The function is also tested on two test sets containing afurther 20 and 10 complexes, respectively. The deficiencies of this type offunction are discussed and it is compared to approaches by other workers.     

Calculation of the solvation free energy of neutral and ionic molecules in diverse solvents

The solvation free energy density (SFED) model was modified to extend its applicability and predictability. The parametrization process was performed with a large, diverse set of solvation free energies that included highly polar and ionic molecules. The mean absolute error for 1200 solvation free energies of the 379 neutral molecules in 9 organic solvents and water was 0.40 kcal/mol, and for 90 hydration free energies of ions was 1.7 kcal/mol. Overall, the calculated solvation free energies of a wide range of solute functional groups in diverse solvents were consistent with experimental data.