QM/MM free energy simulations: recent progress and challenges

Due to the higher computational cost relative to pure molecular mechanical (MM) simulations, hybrid quantum mechanical/molecular mechanical (QM/MM) free energy simulations particularly require a careful consideration of balancing computational cost and accuracy. Here, we review several recent developments in free energy methods most relevant to QM/MM simulations and discuss several topics motivated by these developments using simple but informative examples that involve processes in water. For chemical reactions, we highlight the value of invoking enhanced sampling technique (e.g. replica-exchange) in umbrella sampling calculations and the value of including collective environmental variables (e.g. hydration level) in metadynamics simulations; we also illustrate the sensitivity of string calculations, especially free energy along the path, to various parameters in the computation. Alchemical free energy simulations with a specific thermodynamic cycle are used to probe the effect of including the first solvation shell into the QM region when computing solvation free energies. For cases where high-level QM/MM potential functions are needed, we analyse two different approaches: the QM/MM–MFEP method of Yang and co-workers and perturbative correction to low-level QM/MM free energy results. For the examples analysed here, both approaches seem productive although care needs to be exercised when analysing the perturbative corrections.     

Unconstrained enhanced sampling for free energy calculations of biomolecules: a review

Abstract

Free energy calculations are central to understanding the structure, dynamics and function of biomolecules. Yet insufficient sampling of biomolecular configurations is often regarded as one of the main sources of error. Many enhanced sampling techniques have been developed to address this issue. Notably, enhanced sampling methods based on biasing collective variables (CVs), including the widely used umbrella sampling, adaptive biasing force and metadynamics, have been discussed in a recent excellent review (Abrams and Bussi, Entropy, 2014). Here, we aim to review enhanced sampling methods that do not require predefined system-dependent CVs for biomolecular simulations and as such do not suffer from the hidden energy barrier problem as encountered in the CV-biasing methods. These methods include, but are not limited to, replica exchange/parallel tempering, self-guided molecular/Langevin dynamics, essential energy space random walk and accelerated molecular dynamics. While it is overwhelming to describe all details of each method, we provide a summary of the methods along with the applications and offer our perspectives. We conclude with challenges and prospects of the unconstrained enhanced sampling methods for accurate biomolecular free energy calculations.     

Nonlinear machine learning in simulations of soft and biological materials

Abstract

Interpretable parameterisations of free energy landscapes for soft and biological materials calculated from molecular simulation require the availability of ‘good’ collective variables (CVs) capable of discriminating the metastable states of the system and the barriers between them. If these CVs are coincident with the slow collective modes governing the long-time dynamical evolution, then they also furnish good coordinates in which to perform enhanced sampling to surmount high free energy barriers and efficiently explore and recover the landscape. Non-linear manifold learning techniques provide a means to systematically extract such CVs from molecular simulation trajectories by identifying and extracting low-dimensional manifolds lying latent within the high-dimensional coordinate space. We survey recent advances in data-driven CV discovery and enhanced sampling using non-linear manifold learning, describe the mathematical and theoretical underpinnings of these techniques, and present illustrative examples to molecular folding and colloidal self-assembly. We close with our outlook and perspective on future advances in this rapidly evolving field.     

Putative Active States of a Prototypic G-Protein-Coupled Receptor from Biased Molecular Dynamics

A major current focus of structural work on G-protein-coupled receptors (GPCRs) pertains to the investigation of their active states. However, for virtually all GPCRs, active agonist-bound intermediate states have been difficult to characterize experimentally owing to their higher conformational flexibility, and thus intrinsic instability, as compared to inactive inverse agonist-bound states. In this work, we explored possible activation pathways of the prototypic GPCR bovine rhodopsin by means of biased molecular dynamics simulations. Specifically, we used an explicit atomistic representation of the receptor and its environment, and sampled the conformational transition from the crystal structure of a photoactivated deprotonated state of rhodopsin to the low pH crystal structure of opsin in the presence of 11-trans-retinal, using adiabatic biased molecular dynamics simulations. We then reconstructed the system free-energy landscape along the predetermined transition trajectories using a path collective variable approach based on metadynamics. Our results suggest that the two experimental endpoints of rhodopsin/opsin are connected by at least two different pathways, and that the conformational transition is populated by at least four metastable states of the receptor, characterized by a different amplitude of the outward movement of transmembrane helix 6.     

Computing free energies using nested sampling-based approaches

Abstract

Nested sampling (NS) has emerged as a powerful statistical mechanical sampling technique to compute the partition function of atomic and molecular systems. From the partition function all thermodynamic quantities can be calculated in absolute terms, including absolute free energies and entropies. In this article, we provide a brief overview of NS within a Bayesian context, as well as overviews of how NS is used to compute the partition functions and thermodynamic quantities in the canonical and isothermal-isobaric ensembles. Then we introduce a new scheme, Coupling Parameter Path Nested Sampling, to estimate the free energy difference between two systems with different potential energy functions. The method uses a NS simulation to traverse the same path through phase space as would be covered in traditional coupling parameter-based methods such as thermodynamic integration and perturbation approaches. We demonstrate the new method with two case studies and confirm its accuracy by comparison to conventional methods, including Widom test particle insertion and thermodynamic integration. The proposed method provides a powerful alternative to traditional coupling parameter-based free energy simulation methods.     

Enhanced sampling of molecular dynamics simulation of peptides and proteins by double coupling to thermal bath

Low sampling efficiency in conformational space is the well-known problem for conventional molecular dynamics. It greatly increases the difficulty for molecules to find the transition path to native state, and costs amount of CPU time. To accelerate the sampling, in this paper, we re-couple the critical degrees of freedom in the molecule to environment temperature, like dihedrals in generalized coordinates or nonhydrogen atoms in Cartesian coordinate. After applying to ALA dipeptide model, we find that this modified molecular dynamics greatly enhances the sampling behavior in the conformational space and provides more information about the state-to-state transition, while conventional molecular dynamics fails to do so. Moreover, from the results of 16 independent 100?ns simulations by the new method, it shows that trpzip2 has one-half chances to reach the naive state in all the trajectories, which is greatly higher than conventional molecular dynamics. Such an improvement would provide a potential way for searching the conformational space or predicting the most stable states of peptides and proteins.     

Elucidating the ensemble of functionally-relevant transitions in protein systems with a robotics-inspired method

Background

Many proteins tune their biological function by transitioning between different functional states, effectively acting as dynamic molecular machines. Detailed structural characterization of transition trajectories is central to understanding the relationship between protein dynamics and function. Computational approaches that build on the Molecular Dynamics framework are in principle able to model transition trajectories at great detail but also at considerable computational cost. Methods that delay consideration of dynamics and focus instead on elucidating energetically-credible conformational paths connecting two functionally-relevant structures provide a complementary approach. Effective sampling-based path planning methods originating in robotics have been recently proposed to produce conformational paths. These methods largely model short peptides or address large proteins by simplifying conformational space.

Methods

We propose a robotics-inspired method that connects two given structures of a protein by sampling conformational paths. The method focuses on small- to medium-size proteins, efficiently modeling structural deformations through the use of the molecular fragment replacement technique. In particular, the method grows a tree in conformational space rooted at the start structure, steering the tree to a goal region defined around the goal structure. We investigate various bias schemes over a progress coordinate for balance between coverage of conformational space and progress towards the goal. A geometric projection layer promotes path diversity. A reactive temperature scheme allows sampling of rare paths that cross energy barriers.

Results and conclusions

Experiments are conducted on small- to medium-size proteins of length up to 214 amino acids and with multiple known functionally-relevant states, some of which are more than 13Å apart of each-other. Analysis reveals that the method effectively obtains conformational paths connecting structural states that are significantly different. A detailed analysis on the depth and breadth of the tree suggests that a soft global bias over the progress coordinate enhances sampling and results in higher path diversity. The explicit geometric projection layer that biases the exploration away from over-sampled regions further increases coverage, often improving proximity to the goal by forcing the exploration to find new paths. The reactive temperature scheme is shown effective in increasing path diversity, particularly in difficult structural transitions with known high-energy barriers.

     

Path Similarity Analysis: A Method for Quantifying Macromolecular Pathways

Diverse classes of proteins function through large-scale conformational changes and various sophisticated computational algorithms have been proposed to enhance sampling of these macromolecular transition paths. Because such paths are curves in a high-dimensional space, it has been difficult to quantitatively compare multiple paths, a necessary prerequisite to, for instance, assess the quality of different algorithms. We introduce a method named Path Similarity Analysis (PSA) that enables us to quantify the similarity between two arbitrary paths and extract the atomic-scale determinants responsible for their differences. PSA utilizes the full information available in 3N-dimensional configuration space trajectories by employing the Hausdorff or Fréchet metrics (adopted from computational geometry) to quantify the degree of similarity between piecewise-linear curves. It thus completely avoids relying on projections into low dimensional spaces, as used in traditional approaches. To elucidate the principles of PSA, we quantified the effect of path roughness induced by thermal fluctuations using a toy model system. Using, as an example, the closed-to-open transitions of the enzyme adenylate kinase (AdK) in its substrate-free form, we compared a range of protein transition path-generating algorithms. Molecular dynamics-based dynamic importance sampling (DIMS) MD and targeted MD (TMD) and the purely geometric FRODA (Framework Rigidity Optimized Dynamics Algorithm) were tested along with seven other methods publicly available on servers, including several based on the popular elastic network model (ENM). PSA with clustering revealed that paths produced by a given method are more similar to each other than to those from another method and, for instance, that the ENM-based methods produced relatively similar paths. PSA applied to ensembles of DIMS MD and FRODA trajectories of the conformational transition of diphtheria toxin, a particularly challenging example, showed that the geometry-based FRODA occasionally sampled the pathway space of force field-based DIMS MD. For the AdK transition, the new concept of a Hausdorff-pair map enabled us to extract the molecular structural determinants responsible for differences in pathways, namely a set of conserved salt bridges whose charge-charge interactions are fully modelled in DIMS MD but not in FRODA. PSA has the potential to enhance our understanding of transition path sampling methods, validate them, and to provide a new approach to analyzing conformational transitions.     

Continuous metadynamics in essential coordinates as a tool for free energy modelling of conformational changes

Modelling of conformational changes in biopolymers is one of the greatest challenges of molecular biophysics. Metadynamics is a recently introduced free energy modelling technique that enhances sampling of configurational (e.g. conformational) space within a molecular dynamics simulation. This enhancement is achieved by the addition of a history-dependent bias potential, which drives the system from previously visited regions. Discontinuous metadynamics in the space of essential dynamics eigenvectors (collective motions) has been proposed and tested in conformational change modelling. Here, we present an implementation of two continuous formulations of metadynamics in the essential subspace. The method was performed in a modified version of the molecular dynamics package GROMACS. These implementations were tested on conformational changes in cyclohexane, alanine dipeptide (terminally blocked alanine, Ace-Ala-Nme) and SH3 domain. The results illustrate that metadynamics in the space of essential coordinates can accurately model free energy surfaces associated with conformational changes. Figure The conformational free energy surface of cyclohexane in the space of the two most intensive collective motions.

---

     

(Un)Folding Mechanisms of the FBP28 WW Domain in Explicit Solvent Revealed by Multiple Rare Event Simulation Methods

We report a numerical study of the (un)folding routes of the truncated FBP28 WW domain at ambient conditions using a combination of four advanced rare event molecular simulation techniques. We explore the free energy landscape of the native state, the unfolded state, and possible intermediates, with replica exchange molecular dynamics. Subsequent application of bias-exchange metadynamics yields three tentative unfolding pathways at room temperature. Using these paths to initiate a transition path sampling simulation reveals the existence of two major folding routes, differing in the formation order of the two main hairpins, and in hydrophobic side-chain interactions. Having established that the hairpin strand separation distances can act as reasonable reaction coordinates, we employ metadynamics to compute the unfolding barriers and find that the barrier with the lowest free energy corresponds with the most likely pathway found by transition path sampling. The unfolding barrier at 300 K is ∼17 k_BT ≈ 42 kJ/mol, in agreement with the experimental unfolding rate constant. This work shows that combining several powerful simulation techniques provides a more complete understanding of the kinetic mechanism of protein folding.     

Bayesian Methods for Examining Hardy–Weinberg Equilibrium

Summary .  Testing for Hardy–Weinberg equilibrium is ubiquitous and has traditionally been carried out via frequentist approaches. However, the discreteness of the sample space means that uniformity of  p -values under the null cannot be assumed, with enumeration of all possible counts, conditional on the minor allele count, offering a computationally expensive way of  p -value calibration. In addition, the interpretation of the subsequent  p -values, and choice of significance threshold depends critically on sample size, because equilibrium will always be rejected at conventional levels with large sample sizes. We argue for a Bayesian approach using both Bayes factors, and the examination of posterior distributions. We describe simple conjugate approaches, and methods based on importance sampling Monte Carlo. The former are convenient because they yield closed-form expressions for Bayes factors, which allow their application to a large number of single nucleotide polymorphisms (SNPs), in particular in genome-wide contexts. We also describe straightforward direct sampling methods for examining posterior distributions of parameters of interest. For large numbers of alleles at a locus we resort to Markov chain Monte Carlo. We discuss a number of possibilities for prior specification, and apply the suggested methods to a number of real datasets.     

Transient tertiary structures in tau,an intrinsically disordered protein

An intrinsically disordered protein (IDP) does not have a definite 3D structure, and because of its highly flexible nature it evolves dynamically in very large and diverse regions of the phase space. A standard molecular dynamics run can sample only a limited region of the latter; even though this kind of simulation may be effective in sampling local temporary secondary structures, it is not sufficient to highlight properties that require a larger sampling of the phase space to be detected, like transient tertiary structures. But if the structure of an IDP is dynamically evolved using metadynamics (an algorithm that keeps track of the regions of the phase space already sampled), the system can be forced to wander in a much larger region of the phase space. We have applied this procedure to the simulation of tau, one of the largest totally disordered proteins. Combining the results of the simulation with small-angle X-ray scattering yields a significant improvement in the sampling of the phase space in comparison with standard molecular dynamics, and provides evidence of extended hairpin- and paperclip-like transient tertiary structures of the molecule. The more persistent tertiary pattern is a hairpin folding encompassing part of the N-terminal, the proline-rich domain, the former repeat and a functionally relevant part of the second repeat.     

Unbinding free energy of acetylcholinesterase bound oxime drugs along the gorge pathway from metadynamics‐umbrella sampling investigation

Because of the pivotal role that the nerve enzyme, acetylcholinesterase plays in terminating nerve impulses at cholinergic synapses. Its active site, located deep inside a 20 Å gorge, is a vulnerable target of the lethal organophosphorus compounds. Potent reactivators of the intoxicated enzyme are nucleophiles, such as bispyridinium oxime that binds to the peripheral anionic site and the active site of the enzyme through suitable cation–π interactions. Atomic scale molecular dynamics and free energy calculations in explicit water are used to study unbinding pathways of two oxime drugs (Ortho‐7 and Obidoxime) from the gorge of the enzyme. The role of enzyme‐drug cation–π interactions are explored with the metadynamics simulation. The metadynamics discovered potential of mean force (PMF) of the unbinding events is refined by the umbrella sampling (US) corrections. The bidimensional free energy landscape of the metadynamics runs are further subjected to finite temperature string analysis to obtain the transition tube connecting the minima and bottlenecks of the unbinding pathway. The PMF is also obtained from US simulations using the biasing potential constructed from the transition tube and are found to be consistent with the metadynamics‐US corrected results. Although experimental structural data clearly shows analogous coordination of the two drugs inside the gorge in the bound state, the PMF of the drug trafficking along the gorge pathway point, within an equilibrium free energy context, to a multistep process that differs from one another. Routes, milestones and subtlety toward the unbinding pathway of the two oximes at finite temperature are identified. Proteins 2014; 82:1799–1818. © 2014 Wiley Periodicals, Inc.     

Simulating the free energy of passive membrane permeation for small molecules

Abstract

Computer simulations of passive membrane permeation provide important microscopic insights into the molecular mechanism of this important biological process that are complementary to experimental data. Our review focuses on the main approaches for calculating the free energy, or potential of mean force, for permeation of small molecules through lipid bilayers. The theoretical background for most currently used methods for potential of mean force calculation is described, including particle insertion, thermodynamic integration, umbrella sampling, metadynamics, adaptive biasing force and milestoning. A brief comparison of strengths and weaknesses of the competing approaches is presented. This is followed by a survey of results obtained by the different methods, with special attention to describing the mechanistic insights generated by modelling and illustrating capabilities of the different techniques. We conclude with a discussion of recent advances and future directions in modelling membrane permeation, including latest methodological enhancements, consideration of multiple slow variables and memory effects.     

Prediction of individual long-term outcomes in smoking cessation trials using frailty models

In smoking cessation clinical trials, subjects commonly receive treatment and report daily cigarette consumption over a period of several weeks. Although the outcome at the end of this period is an important indicator of treatment success, substantial uncertainty remains on how an individual's smoking behavior will evolve over time. Therefore it is of interest to predict long-term smoking cessation success based on short-term clinical observations. We develop a Bayesian method for prediction, based on a cure-mixture frailty model we proposed earlier, that describes the process of transition between abstinence and smoking. Specifically we propose a two-stage prediction algorithm that first uses importance sampling to generate subject-specific frailties from their posterior distributions conditional on the observed data, then samples predicted future smoking behavior trajectories from the estimated model parameters and sampled frailties. We apply the method to data from two randomized smoking cessation trials comparing bupropion to placebo. Comparisons of actual smoking status at one year with predictions from our model and from a variety of empirical methods suggest that our method gives excellent predictions.     

Toward a detailed understanding of search trajectories in fragment assembly approaches to protein structure prediction

Energy functions, fragment libraries, and search methods constitute three key components of fragment‐assembly methods for protein structure prediction, which are all crucial for their ability to generate high‐accuracy predictions. All of these components are tightly coupled; efficient searching becomes more important as the quality of fragment libraries decreases. Given these relationships, there is currently a poor understanding of the strengths and weaknesses of the sampling approaches currently used in fragment‐assembly techniques. Here, we determine how the performance of search techniques can be assessed in a meaningful manner, given the above problems. We describe a set of techniques that aim to reduce the impact of the energy function, and assess exploration in view of the search space defined by a given fragment library. We illustrate our approach using Rosetta and EdaFold, and show how certain features of these methods encourage or limit conformational exploration. We demonstrate that individual trajectories of Rosetta are susceptible to local minima in the energy landscape, and that this can be linked to non‐uniform sampling across the protein chain. We show that EdaFold's novel approach can help balance broad exploration with locating good low‐energy conformations. This occurs through two mechanisms which cannot be readily differentiated using standard performance measures: exclusion of false minima, followed by an increasingly focused search in low‐energy regions of conformational space. Measures such as ours can be helpful in characterizing new fragment‐based methods in terms of the quality of conformational exploration realized. Proteins 2016; 84:411–426. © 2016 The Authors Proteins: Structure, Function, and Bioinformatics Published by Wiley Periodicals, Inc.     

Predicting novel disease mutations in the cardiac sodium channel

BackgroundVoltage-gated sodium channels Nav1.x mediate the rising phase of action potential in excitable cells. Variations in gene SCN5A, which encodes the hNav1.5 channel, are associated with arrhythmias and other heart diseases. About 1,400 SCN5A variants are listed in public databases, but for more than 30% of these the clinical significance is unknown and can currently only be derived by bioinformatics approaches.Methods and resultsWe used the ClinVar, SwissVar, Humsavar, gnomAD, and Ensembl databases to assemble a dataset of 1392 hNav1.5 variants (370 pathogenic variants, 602 benign variants and 420 variants of uncertain significance) as well as a dataset of 1766 damaging variants in 20 human sodium and calcium channel paralogs. Twelve in silico tools were tested for their ability to predict damaging mutations in hNav1.5. The best performing tool, MutPred, correctly predicted 93% of damaging variants in our hNav1.5 dataset. Among the 86 hNav1.5 variants for which electrophysiological data are also available, MutPred correctly predicted 82% of damaging variants. In the subset of 420 uncharacterized hNav1.5 variants MutPred predicted 196 new pathogenic variants. Among these, 74 variants are also annotated as damaging in at least one hNav1.5 paralog.ConclusionsUsing a combination of sequence-based bioinformatics techniques and paralogous annotation we have substantially expanded the knowledge on disease variants in the cardiac sodium channel and assigned a pathogenic status to a number of mutations that so far have been described as variants of uncertain significance. A list of reclassified hNav1.5 variants and their properties is provided.