Optimal intermediates in staged free energy calculations

Implementation of Telescoping Boxes in Adaptive Steered Molecular Dynamics

Long-time dynamical processes, such as those involving protein unfolding and ligand interactions, can be accelerated and realized through steered molecular dynamics (SMD). The challenge has been the extraction of information from such simulations that generalize for complex nonequilibrium processes. The use of Jarzynski's equality opened the possibility of determining the free energy along the steered coordinate, but sampling over the nonequilibrium trajectories is slow to converge. Adaptive steered molecular dynamics (ASMD) and other related techniques have been introduced to overcome this challenge through the use of stages. Here, we take advantage of these stages to address the numerical cost that arises from the required use of very large solvent boxes. We introduce telescoping box schemes within adaptive steered molecular dynamics (ASMD) in which we adjust the solvent box between stages and thereby vary (and optimize) the required number of solvent molecules. We have benchmarked the method on a relatively long α-helical peptide, Ala₃₀, with respect to the potential of mean force and hydrogen bonds. We show that the use of telescoping boxes introduces little numerical error while significantly reducing the computational cost.

Computational Tools for Accurate Binding Free-Energy Prediction

A quantitative thermodynamic understanding of the noncovalent association of (bio)molecules is of central importance in molecular life sciences. An important quantity characterizing (bio)molecular association is the binding affinity or absolute binding free energy. In recent years, the computational prediction of absolute binding free energies has evolved considerably in terms of accuracy, computational speed, and user-friendliness. In this chapter, we first give an overview of how absolute free energies are defined and how they can be determined with computational means. We proceed with an outline of the theoretical basis of the two most reliable methods, potential of mean force, and double decoupling calculations. In particular, we describe how the sampling problem can be alleviated by application of restraints. Finally, we provide step-by-step instructions of how to set up corresponding molecular simulations with a commonly employed molecular dynamics simulation engine.

The accelerated weight histogram method for alchemical free energy calculations

The accelerated weight histogram method is an enhanced sampling technique used to explore free energy landscapes by applying an adaptive bias. The method is general and easy to extend. Herein, we show how it can be used to efficiently sample alchemical transformations, commonly used for, e.g., solvation and binding free energy calculations. We present calculations and convergence of the hydration free energy of testosterone, representing drug-like molecules. We also include methane and ethanol to validate the results. The protocol is easy to use, does not require a careful choice of parameters, and scales well to accessible resources, and the results converge at least as quickly as when using conventional methods. One benefit of the method is that it can easily be combined with other reaction coordinates, such as intermolecular distances.

Skewed thermodynamic geometry and optimal free energy estimation

Free energy differences are a central quantity of interest in physics, chemistry, and biology. We develop design principles that improve the precision and accuracy of free energy estimators, which have potential applications to screening for targeted drug discovery. Specifically, by exploiting the connection between the work statistics of time-reversed protocol pairs, we develop near-equilibrium approximations for moments of the excess work and analyze the dominant contributions to the precision and accuracy of standard nonequilibrium free-energy estimators. Within linear response, minimum-dissipation protocols follow the geodesics of the Riemannian metric induced by the Stokes friction tensor. We find that the next-order contribution arises from the rank-3 supra-Stokes tensor that skews the geometric structure such that minimum-dissipation protocols follow the geodesics of a generalized cubic Finsler metric. Thus, near equilibrium, the supra-Stokes tensor determines the leading-order contribution to the bias of bidirectional free-energy estimators.

Extrapolation and interpolation strategies for efficiently estimating structural observables as a function of temperature and density

Thermodynamic extrapolation has previously been used to predict arbitrary structural observables in molecular simulations at temperatures (or relative chemical potentials in open-system mixtures) different from those at which the simulation was performed. This greatly reduces the computational cost in mapping out phase and structural transitions. In this work, we explore the limitations and accuracy of thermodynamic extrapolation applied to water, where qualitative shifts from anomalous to simple-fluid-like behavior are manifested through shifts in the liquid structure that occur as a function of both temperature and density. We present formulas for extrapolating in volume for canonical ensembles and demonstrate that linear extrapolations of water's structural properties are only accurate over a limited density range. On the other hand, linear extrapolation in temperature can be accurate across the entire liquid state. We contrast these extrapolations with classical perturbation theory techniques, which are more conservative and slowly converging. Indeed, we show that such behavior is expected by demonstrating exact relationships between extrapolation of free energies and well-known techniques to predict free energy differences. An ideal gas in an external field is also studied to more clearly explain these results for a toy system with fully analytical solutions. We also present a recursive interpolation strategy for predicting arbitrary structural properties of molecular fluids over a predefined range of state conditions, demonstrating its success in mapping qualitative shifts in water structure with density.

Robust Free Energy Perturbation Protocols for Creating Molecules in Solution

Accurate methods to estimate free energies play an important role for studying diverse condensed-phase problems in chemistry and biochemistry. The most common methods used in conjunction with molecular dynamics (MD) and Monte Carlo statistical mechanics (MC) simulations are free energy perturbation (FEP) and thermodynamic integration (TI). For common applications featuring the conversion of one molecule to another, simulations are run in stages or multiple "λ-windows" to promote convergence of the results. For computation of absolute free energies of solvation or binding, calculations are needed in which the solute is typically annihilated in the solvent and in the complex. The present work addresses identification of optimal protocols for such calculations, specifically, the creation/annihilation of organic molecules in aqueous solution. As is common practice, decoupling of the perturbations for electrostatic and Lennard-Jones interactions was performed. Consistent with earlier reports, FEP calculations for molecular creations are much more efficient, while annihilations require many more windows and may converge to incorrect values. Strikingly, we find that as few as four windows may be adequate for creation calculations for solutes ranging from argon to ethylbenzene. For a larger druglike molecule, MIF180, which contains 22 non-hydrogen atoms and three rotatable bonds, 10 creation windows are found to be adequate to yield the correct free energy of hydration. Convergence is impeded with procedures that use any sampling in the annihilation direction, and there is no need for postprocessing methods such as the Bennett acceptance ratio (BAR).

Molecular simulation as a computational pharmaceutics tool to predict drug solubility, solubilization processes and partitioning

In this review we will discuss how computational methods, and in particular classical molecular dynamics simulations, can be used to calculate solubility of pharmaceutically relevant molecules and systems. To the extent possible, we focus on the non-technical details of these calculations, and try to show also the added value of a more thorough and detailed understanding of the solubilization process obtained by using computational simulations. Although the main focus is on classical molecular dynamics simulations, we also provide the reader with some insights into other computational techniques, such as the COSMO-method, and also discuss Flory-Huggins theory and solubility parameters. We hope that this review will serve as a valuable starting point for any pharmaceutical researcher, who has not yet fully explored the possibilities offered by computational approaches to solubility calculations.

Chemical potential calculations in non-homogeneous liquids

The numerical computation of chemical potential in dense non-homogeneous fluids is a key problem in the study of confined fluid thermodynamics. To this day, several methods have been proposed; however, there is still need for a robust technique, capable of obtaining accurate estimates at large average densities. A widely established technique is the Widom insertion method, which computes the chemical potential by sampling the energy of insertion of a test particle. Non-homogeneity is accounted for by assigning a density dependent weight to the insertion points. However, in dense systems, the poor sampling of the insertion energy is a source of inefficiency, hampering a reliable convergence. We have recently presented a new technique for the chemical potential calculation in homogeneous fluids. This novel method enhances the sampling of the insertion energy via well-tempered metadynamics, reaching accurate estimates at very large densities. In this paper, we extend the technique to the case of non-homogeneous fluids. The method is successfully tested on a confined Lennard-Jones fluid. In particular, we show that, thanks to the improved sampling, our technique does not suffer from a systematic error that affects the classic Widom method for non-homogeneous fluids, providing a precise and accurate result.

Statistical Analysis on the Performance of Molecular Mechanics Poisson-Boltzmann Surface Area versus Absolute Binding Free Energy Calculations: Bromodomains as a Case Study

Binding free energy calculations that make use of alchemical pathways are becoming increasingly feasible thanks to advances in hardware and algorithms. Although relative binding free energy (RBFE) calculations are starting to find widespread use, absolute binding free energy (ABFE) calculations are still being explored mainly in academic settings due to the high computational requirements and still uncertain predictive value. However, in some drug design scenarios, RBFE calculations are not applicable and ABFE calculations could provide an alternative. Computationally cheaper end-point calculations in implicit solvent, such as molecular mechanics Poisson-Boltzmann surface area (MMPBSA) calculations, could too be used if one is primarily interested in a relative ranking of affinities. Here, we compare MMPBSA calculations to previously performed absolute alchemical free energy calculations in their ability to correlate with experimental binding free energies for three sets of bromodomain-inhibitor pairs. Different MMPBSA approaches have been considered, including a standard single-trajectory protocol, a protocol that includes a binding entropy estimate, and protocols that take into account the ligand hydration shell. Despite the improvements observed with the latter two MMPBSA approaches, ABFE calculations were found to be overall superior in obtaining correlation with experimental affinities for the test cases considered. A difference in weighted average Pearson ([Formula: see text]) and Spearman ([Formula: see text]) correlations of 0.25 and 0.31 was observed when using a standard single-trajectory MMPBSA setup ([Formula: see text] = 0.64 and [Formula: see text] = 0.66 for ABFE; [Formula: see text] = 0.39 and [Formula: see text] = 0.35 for MMPBSA). The best performing MMPBSA protocols returned weighted average Pearson and Spearman correlations that were about 0.1 inferior to ABFE calculations: [Formula: see text] = 0.55 and [Formula: see text] = 0.56 when including an entropy estimate, and [Formula: see text] = 0.53 and [Formula: see text] = 0.55 when including explicit water molecules. Overall, the study suggests that ABFE calculations are indeed the more accurate approach, yet there is also value in MMPBSA calculations considering the lower compute requirements, and if agreement to experimental affinities in absolute terms is not of interest. Moreover, for the specific protein-ligand systems considered in this study, we find that including an explicit ligand hydration shell or a binding entropy estimate in the MMPBSA calculations resulted in significant performance improvements at a negligible computational cost.

Constrained Unfolding of a Helical Peptide: Implicit versus Explicit Solvents

Steered Molecular Dynamics (SMD) has been seen to provide the potential of mean force (PMF) along a peptide unfolding pathway effectively but at significant computational cost, particularly in all-atom solvents. Adaptive steered molecular dynamics (ASMD) has been seen to provide a significant computational advantage by limiting the spread of the trajectories in a staged approach. The contraction of the trajectories at the end of each stage can be performed by taking a structure whose nonequilibrium work is closest to the Jarzynski average (in naive ASMD) or by relaxing the trajectories under a no-work condition (in full-relaxation ASMD--namely, FR-ASMD). Both approaches have been used to determine the energetics and hydrogen-bonding structure along the pathway for unfolding of a benchmark peptide initially constrained as an α-helix in a water environment. The energetics are quite different to those in vacuum, but are found to be similar between implicit and explicit solvents. Surprisingly, the hydrogen-bonding pathways are also similar in the implicit and explicit solvents despite the fact that the solvent contact plays an important role in opening the helix.

An introduction to best practices in free energy calculations

Free energy calculations are extremely useful for investigating small-molecule biophysical properties such as protein-ligand binding affinities and partition coefficients. However, these calculations are also notoriously difficult to implement correctly. In this chapter, we review standard methods for computing free energy via simulation, discussing current best practices and examining potential pitfalls for computational researchers performing them for the first time. We include a variety of examples and tips for how to set up and conduct these calculations, including applications to relative binding affinities and small-molecule solvation free energies.

Good practices in free-energy calculations

As access to computational resources continues to increase, free-energy calculations have emerged as a powerful tool that can play a predictive role in a wide range of research areas. Yet, the reliability of these calculations can often be improved significantly if a number of precepts, or good practices, are followed. Although the theory upon which these good practices rely has largely been known for many years, it is often overlooked or simply ignored. In other cases, the theoretical developments are too recent for their potential to be fully grasped and merged into popular platforms for the computation of free-energy differences. In this contribution, the current best practices for carrying out free-energy calculations using free energy perturbation and nonequilibrium work methods are discussed, demonstrating that at little to no additional cost, free-energy estimates could be markedly improved and bounded by meaningful error estimates. Monitoring the probability distributions that underlie the transformation between the states of interest, performing the calculation bidirectionally, stratifying the reaction pathway, and choosing the most appropriate paradigms and algorithms for transforming between states offer significant gains in both accuracy and precision.

Minimizing thermodynamic length to select intermediate states for free-energy calculations and replica-exchange simulations

In computational thermodynamics, a sequence of intermediate states is commonly introduced to connect two equilibrium states. We consider two cases where the choice of intermediate states is particularly important: minimizing statistical error in free-energy difference calculations and maximizing average acceptance probabilities in replica-exchange simulations. We derive bounds for these quantities in terms of the thermodynamic distance between the intermediates, and show that in both cases the intermediates should be chosen as equidistant points along a geodesic connecting the end states.

Optimized expanded ensembles for simulations involving molecular insertions and deletions. II. Open systems

In the Grand Canonical, osmotic, and Gibbs ensembles, chemical potential equilibrium is attained via transfers of molecules between the system and either a reservoir or another subsystem. In this work, the expanded ensemble (EXE) methods described in part I [F. A. Escobedo and F. J. Martinez-Veracoechea, J. Chem. Phys. 127, 174103 (2007)] of this series are extended to these ensembles to overcome the difficulties associated with implementing such whole-molecule transfers. In EXE, such moves occur via a target molecule that undergoes transitions through a number of intermediate coupling states. To minimize the tunneling time between the fully coupled and fully decoupled states, the intermediate states could be either: (i) sampled with an optimal frequency distribution (the sampling problem) or (ii) selected with an optimal spacing distribution (staging problem). The sampling issue is addressed by determining the biasing weights that would allow generating an optimal ensemble; discretized versions of this algorithm (well suited for small number of coupling stages) are also presented. The staging problem is addressed by selecting the intermediate stages in such a way that a flat histogram is the optimized ensemble. The validity of the advocated methods is demonstrated by their application to two model problems, the solvation of large hard spheres into a fluid of small and large spheres, and the vapor-liquid equilibrium of a chain system.

Comparison of free energy methods for molecular systems

We present a detailed comparison of computational efficiency and precision for several free energy difference (DeltaF) methods. The analysis includes both equilibrium and nonequilibrium approaches, and distinguishes between unidirectional and bidirectional methodologies. We are primarily interested in comparing two recently proposed approaches, adaptive integration, and single-ensemble path sampling to more established methodologies. As test cases, we study relative solvation free energies of large changes to the size or charge of a Lennard-Jones particle in explicit water. The results show that, for the systems used in this study, both adaptive integration and path sampling offer unique advantages over the more traditional approaches. Specifically, adaptive integration is found to provide very precise long-simulation DeltaF estimates as compared to other methods used in this report, while also offering rapid estimation of DeltaF. The results demonstrate that the adaptive integration approach is the best overall method for the systems studied here. The single-ensemble path sampling approach is found to be superior to ordinary Jarzynski averaging for the unidirectional, "fast-growth" nonequilibrium case. Closer examination of the path sampling approach on a two-dimensional system suggests it may be the overall method of choice when conformational sampling barriers are high. However, it appears that the free energy landscapes for the systems used in this study have rather modest configurational sampling barriers.

Phase-space overlap measures. I. Fail-safe bias detection in free energies calculated by molecular simulation

We consider ways to quantify the overlap of the parts of phase space important to two systems, labeled A and B. Of interest is how much of the A-important phase space lies in that important to B, and how much of B lies in A. Two measures are proposed. The first considers four total-energy distributions, formed from all combinations made by tabulating either the A-system or the B-system energy when sampling either the A or B system. Measures for A in B and B in A are given by two overlap integrals defined on pairs of these distributions. The second measure is based on information theory, and defines two relative entropies which are conveniently expressed in terms of the dissipated work for free-energy perturbation (FEP) calculations in the A-->B and B-->A directions, respectively. Phase-space overlap is an important consideration in the performance of free-energy calculations. To demonstrate this connection, we examine bias in FEP calculations applied to a system of independent particles in a harmonic potential. Systems are selected to represent a range of overlap situations, including extreme subset, subset, partial overlap, and nonoverlap. The magnitude and symmetry of the bias (A-->B vs B-->A) are shown to correlate well with the overlap, and consequently with the overlap measures. The relative entropies are used to scale the amount of sampling to obtain a universal bias curve. This result leads to develop a simple heuristic that can be applied to determine whether a work-based free-energy measurement is free of bias. The heuristic is based in part on the measured free energy, but we argue that it is fail-safe inasmuch as any bias in the measurement will not promote a false indication of accuracy.

Model for small-sample bias of free-energy calculations applied to Gaussian-distributed nonequilibrium work measurements

We present a model for the bias of free-energy differences when determined using the nonequilibrium work (NEW) formalism due to Jarzynski. Input to the model is the distribution of work values underlying the NEW calculation, and the bias is estimated by assuming that all of the inaccuracy is incurred by failure to sample work values beyond a specific point in the tail of the distribution. The model is formulated considering both small- and large-sample NEW calculations. It is then applied to the study of bias for cases in which the work distribution is Gaussian. The model is shown to give an excellent description of the bias in situations where the bias is a more significant source of error than the sample variance. A scaling law is presented and it is shown that the bias as a function of sampling can be reduced to a single universal curve, approximately valid for all Gaussian work distributions. This result is used to formulate a simple criterion that can be applied to test if a NEW calculation is effectively free of bias. The criterion is shown to be effective even though it uses the measured (and perhaps biased) free energy as an input.

Comparison of efficiency and bias of free energies computed by exponential averaging, the Bennett acceptance ratio, and thermodynamic integration

Recent work has demonstrated the Bennett acceptance ratio method is the best asymptotically unbiased method for determining the equilibrium free energy between two end states given work distributions collected from either equilibrium and nonequilibrium data. However, it is still not clear what the practical advantage of this acceptance ratio method is over other common methods in atomistic simulations. In this study, we first review theoretical estimates of the bias and variance of exponential averaging (EXP), thermodynamic integration (TI), and the Bennett acceptance ratio (BAR). In the process, we present a new simple scheme for computing the variance and bias of many estimators, and demonstrate the connections between BAR and the weighted histogram analysis method. Next, a series of analytically solvable toy problems is examined to shed more light on the relative performance in terms of the bias and efficiency of these three methods. Interestingly, it is impossible to conclusively identify a "best" method for calculating the free energy, as each of the three methods performs more efficiently than the others in at least one situation examined in these toy problems. Finally, sample problems of the insertion/deletion of both a Lennard-Jones particle and a much larger molecule in TIP3P water are examined by these three methods. In all tests of atomistic systems, free energies obtained with BAR have significantly lower bias and smaller variance than when using EXP or TI, especially when the overlap in phase space between end states is small. For example, BAR can extract as much information from multiple fast, far-from-equilibrium simulations as from fewer simulations near equilibrium, which EXP cannot. Although TI and sometimes even EXP can be somewhat more efficient in idealized toy problems, in the realistic atomistic situations tested in this paper, BAR is significantly more efficient than all other methods.

Low-Temperature Vapor−Liquid Equilibria from Parallelized Molecular Dynamics Simulations. Application to 1- and 2-Methylnaphthalene

A parallelized sampling version of the Gibbs Ensemble (Mol. Phys. 2000, 98, 1887) has been implemented to predict low-temperature vapor-liquid equilibria of 1- and 2-methylnaphthalene modeled by anisotropic united atom potentials. The simulation were performed at the low temperature of 364.2 K at which common direct simulation methods fail due to particle transfer problems. The simulation results are compared with published results obtained from the Gibbs-Duhem integration method and with experimental data. Both methods are compared and discussed in terms of computational efficiency and with respect to their future use at other thermodynamic conditions.

Tools for channels: moving towards molecular calculations of gating and permeation in ion channel biophysics

Recent X-ray structures of voltage gated potassium channels provide an exciting opportunity to connect molecular structures with measured biological function. Two of the most important connections for these channels are: first, to the molecular basis behind selectivity and the associated free energy profile underlying ionic current flow and, second, to a true molecular understanding of the large-scale conformational transitions that underlie voltage dependent gating. But, existing computational tools need to be further developed to reach these goals. In this contribution to the symposia on sampling methods we outline our dynamic importance sampling method for sampling large-scale conformational transitions as well as our studies with non-equilibrium work events and equilibrium overlap sampling (OS) methods for sampling events related to the calculation of relative free energies.

Equilibrium free energies from nonequilibrium measurements using maximum-likelihood methods

We present a maximum likelihood argument for the Bennett acceptance ratio method, and derive a simple formula for the variance of free energy estimates generated using this method. This derivation of the acceptance ratio method, using a form of logistic regression, a common statistical technique, allows us to shed additional light on the underlying physical and statistical properties of the method. For example, we demonstrate that the acceptance ratio method yields the lowest variance for any estimator of the free energy which is unbiased in the limit of large numbers of measurements.

Variational formula for the free energy based on incomplete sampling in a molecular simulation

Finite sampling in free-energy perturbation (FEP) calculations by molecular simulation leads to reproducible systematic errors, with sign shown to depend (in a known way) only on which system governs sampling in the simulation. Thus the result of a FEP calculation can be used as a bound on the true free energy. This inequality is of a wholly different nature from established forms such as the Gibbs-Bogoliubov inequality or the second law, in that its origins relate to the performance of a molecular simulation. If one can identify a suitable reference system having a free energy known as a function of some defining parameter, variational schemes based on the finite-sampling inequalities can be implemented. This idea is demonstrated through calculation of the free energy of a hard-sphere solid by perturbing from harmonic references and of a hard-sphere fluid by perturbing from infinitely polydisperse references. The tightness of the bounds increases with the amount of sampling in the simulation and correlates with the entropy difference between the target and reference systems. The bounds are tightest near the point where the entropy difference is least.

Improving the efficiency and reliability of free energy perturbation calculations using overlap sampling methods

A challenge in free energy calculation for complex molecular systems by computer simulation is to obtain a reliable estimate within feasible computational time. In this study, we suggest an answer to this challenge by exploring a simple method, overlap sampling (OS), for producing reliable free-energy results in an efficient way. The formalism of the OS method is based on ensuring sampling of important overlapping phase space during perturbation calculations. This technique samples both forward and reverse free energy perturbation (FEP) to improve the free-energy calculation. It considers the asymmetry of the FEP calculation and features an ability to optimize both the precision and the accuracy of the measurement without affecting the simulation process itself. The OS method is tested at two optimization levels: no optimization (simple OS), and full optimization (equivalent to Bennett's method), and compared to conventional FEP techniques, including the widely used direct FEP averaging method, on three alchemical mutation systems: (a) an anion transformation in water solution, (b) mutation between methanol and ethane, and (c) alchemical change of an adenosine molecule. It is consistently shown that the reliability of free-energy estimates can be greatly improved using the OS techniques at both optimization levels, while the performance of Bennett's method is particularly striking. In addition, the efficiency of a calculation can be significantly improved because the method is able to (a) converge to the right answer quickly, and (b) work for large perturbations. The basic two-stage OS method can be extended to admit additional stages, if needed. We suggest that the OS method can be used as a general perturbation technique for computing free energy differences in molecular simulations.

Self-referential method for calculation of the free energy of crystals by Monte Carlo simulation

We propose a Monte Carlo simulation method for the evaluation of free energies in crystalline systems. In principle, the method involves evaluating the free-energy difference between systems of N molecules and 2N molecules. This difference, coupled with the assumption that the free energy is extensive and thus proportional to N, provides sufficient information to obtain the absolute free energy of the crystal. The approach to doubling the system size does not involve insertion or removal of molecules in the system. Instead, the configurations of the molecules are expressed in terms of the normal-mode coordinates of a harmonic lattice. By decoupling certain of these coordinates from the molecule configurations, we obtain a transformation that in effect yields the system-size doubling. The method is examined via application to a system of hard rods in one dimension. This simple model is considered principally because of the availability of an analytic solution for its free energy, which permits accurate testing of the performance and correctness of the proposed method. In using the hard-rod model we also avoid other complications related to treatment of the temperature, and application of normal-mode coordinate decoupling in higher dimensions. The proposed method is shown to be able to provide good results for the free-energy calculation, but further development will be needed before it can be considered practical for general-purpose use.