Developing an Implicit Solvation Machine Learning Model for Molecular Simulations of Ionic Media.

Molecular dynamics (MD) simulations of biophysical systems require accurate modeling of their native environment, i.e., aqueous ionic solution, as it critically impacts the structure and function of biomolecules. On the other hand, the models should be computationally efficient to enable simulations of large spatiotemporal scales. Here, we present the deep implicit solvation model for sodium chloride solutions that satisfies both requirements. Owing to the use of the neural network potential, the model can capture the many-body potential of mean force, while the implicit water treatment renders the model inexpensive. We demonstrate our approach first for pure ionic solutions with concentrations ranging from physiological to 2 M. We then extend the model to capture the effective ion interactions in the vicinity and far away from a DNA molecule. In both cases, the structural properties are in good agreement with all-atom MD, showcasing a general methodology for the efficient and accurate modeling of ionic media.

Scalable Bayesian Uncertainty Quantification for Neural Network Potentials: Promise and Pitfalls.

Neural network (NN) potentials promise highly accurate molecular dynamics (MD) simulations within the computational complexity of classical MD force fields. However, when applied outside their training domain, NN potential predictions can be inaccurate, increasing the need for Uncertainty Quantification (UQ). Bayesian modeling provides the mathematical framework for UQ, but classical Bayesian methods based on Markov chain Monte Carlo (MCMC) are computationally intractable for NN potentials. By training graph NN potentials for coarse-grained systems of liquid water and alanine dipeptide, we demonstrate here that scalable Bayesian UQ via stochastic gradient MCMC (SG-MCMC) yields reliable uncertainty estimates for MD observables. We show that cold posteriors can reduce the required training data size and that for reliable UQ, multiple Markov chains are needed. Additionally, we find that SG-MCMC and the Deep Ensemble method achieve comparable results, despite shorter training and less hyperparameter tuning of the latter. We show that both methods can capture aleatoric and epistemic uncertainty reliably, but not systematic uncertainty, which needs to be minimized by adequate modeling to obtain accurate credible intervals for MD observables. Our results represent a step toward accurate UQ that is of vital importance for trustworthy NN potential-based MD simulations required for decision-making in practice.

Deep coarse-grained potentials via relative entropy minimization.

Neural network (NN) potentials are a natural choice for coarse-grained (CG) models. Their many-body capacity allows highly accurate approximations of the potential of mean force, promising CG simulations of unprecedented accuracy. CG NN potentials trained bottom-up via force matching (FM), however, suffer from finite data effects: They rely on prior potentials for physically sound predictions outside the training data domain, and the corresponding free energy surface is sensitive to errors in the transition regions. The standard alternative to FM for classical potentials is relative entropy (RE) minimization, which has not yet been applied to NN potentials. In this work, we demonstrate, for benchmark problems of liquid water and alanine dipeptide, that RE training is more data efficient, due to accessing the CG distribution during training, resulting in improved free energy surfaces and reduced sensitivity to prior potentials. In addition, RE learns to correct time integration errors, allowing larger time steps in CG molecular dynamics simulation, while maintaining accuracy. Thus, our findings support the use of training objectives beyond FM, as a promising direction for improving CG NN potential's accuracy and reliability.

Accelerated Simulations of Molecular Systems through Learning of Effective Dynamics.

Simulations are vital for understanding and predicting the evolution of complex molecular systems. However, despite advances in algorithms and special purpose hardware, accessing the time scales necessary to capture the structural evolution of biomolecules remains a daunting task. In this work, we present a novel framework to advance simulation time scales by up to 3 orders of magnitude by learning the effective dynamics (LED) of molecular systems. LED augments the equation-free methodology by employing a probabilistic mapping between coarse and fine scales using mixture density network (MDN) autoencoders and evolves the non-Markovian latent dynamics using long short-term memory MDNs. We demonstrate the effectiveness of LED in the Müller-Brown potential, the Trp cage protein, and the alanine dipeptide. LED identifies explainable reduced-order representations, i.e., collective variables, and can generate, at any instant, all-atom molecular trajectories consistent with the collective variables. We believe that the proposed framework provides a dramatic increase to simulation capabilities and opens new horizons for the effective modeling of complex molecular systems.

Learning neural network potentials from experimental data via Differentiable Trajectory Reweighting.

In molecular dynamics (MD), neural network (NN) potentials trained bottom-up on quantum mechanical data have seen tremendous success recently. Top-down approaches that learn NN potentials directly from experimental data have received less attention, typically facing numerical and computational challenges when backpropagating through MD simulations. We present the Differentiable Trajectory Reweighting (DiffTRe) method, which bypasses differentiation through the MD simulation for time-independent observables. Leveraging thermodynamic perturbation theory, we avoid exploding gradients and achieve around 2 orders of magnitude speed-up in gradient computation for top-down learning. We show effectiveness of DiffTRe in learning NN potentials for an atomistic model of diamond and a coarse-grained model of water based on diverse experimental observables including thermodynamic, structural and mechanical properties. Importantly, DiffTRe also generalizes bottom-up structural coarse-graining methods such as iterative Boltzmann inversion to arbitrary potentials. The presented method constitutes an important milestone towards enriching NN potentials with experimental data, particularly when accurate bottom-up data is unavailable.

Tuning the Dielectric Response of Water in Nanoconfinement through Surface Wettability.

The tunable polarity of water can be exploited in emerging technologies including catalysis, gas storage, and green chemistry. Recent experimental and theoretical studies have shown that water can be rendered into an effectively apolar solvent under nanoconfinement. We furthermore demonstrate, through molecular simulations, that the static dielectric constant of water can be modified by changing the wettability of the confining material. We find the out-of-plane dielectric response to be highly sensitive to the level of confinement and can be reduced up to 40× , in accordance with experimental data. By altering the surface wettability from superhydrophilic to superhydrophobic, we observe a 36% increase for the out-of-plane and a 31% decrease for the in-plane dielectric constants. Our findings demonstrate the feasibility of tunable water polarity, a phenomenon with great potential for scientific and technological impact.

SWINGER: a clustering algorithm for concurrent coupling of atomistic and supramolecular liquids.

In this contribution, we review recent developments and applications of a dynamic clustering algorithm SWINGER tailored for the multiscale molecular simulations of biomolecular systems. The algorithm on-the-fly redistributes solvent molecules among supramolecular clusters. In particular, we focus on its applications in combination with the adaptive resolution scheme, which concurrently couples atomistic and coarse-grained molecular representations. We showcase the versatility of our multiscale approach on a few applications to biomolecular systems coupling atomistic and supramolecular water models such as the well-established MARTINI and dissipative particle dynamics models and provide an outlook for future work.

Bayesian selection for coarse-grained models of liquid water.

The necessity for accurate and computationally efficient representations of water in atomistic simulations that can span biologically relevant timescales has born the necessity of coarse-grained (CG) modeling. Despite numerous advances, CG water models rely mostly on a-priori specified assumptions. How these assumptions affect the model accuracy, efficiency, and in particular transferability, has not been systematically investigated. Here we propose a data driven comparison and selection for CG water models through a Hierarchical Bayesian framework. We examine CG water models that differ in their level of coarse-graining, structure, and number of interaction sites. We find that the importance of electrostatic interactions for the physical system under consideration is a dominant criterion for the model selection. Multi-site models are favored, unless the effects of water in electrostatic screening are not relevant, in which case the single site model is preferred due to its computational savings. The charge distribution is found to play an important role in the multi-site model's accuracy while the flexibility of the bonds/angles may only slightly improve the models. Furthermore, we find significant variations in the computational cost of these models. We present a data informed rationale for the selection of CG water models and provide guidance for future water model designs.

Open-Boundary Molecular Dynamics of a DNA Molecule in a Hybrid Explicit/Implicit Salt Solution.

The composition and electrolyte concentration of the aqueous bathing environment have important consequences for many biological processes and can profoundly affect the behavior of biomolecules. Nevertheless, because of computational limitations, many molecular simulations of biophysical systems can be performed only at specific ionic conditions: either at nominally zero salt concentration, i.e., including only counterions enforcing the system's electroneutrality, or at excessive salt concentrations. Here, we introduce an efficient molecular dynamics simulation approach for an atomistic DNA molecule at realistic physiological ionic conditions. The simulations are performed by employing the open-boundary molecular dynamics method that allows for simulation of open systems that can exchange mass and linear momentum with the environment. In our open-boundary molecular dynamics approach, the computational burden is drastically alleviated by embedding the DNA molecule in a mixed explicit/implicit salt-bathing solution. In the explicit domain, the water molecules and ions are both overtly present in the system, whereas in the implicit water domain, only the ions are explicitly present and the water is described as a continuous dielectric medium. Water molecules are inserted and deleted into/from the system in the intermediate buffer domain that acts as a water reservoir to the explicit domain, with both water molecules and ions free to enter or leave the explicit domain. Our approach is general and allows for efficient molecular simulations of biomolecules solvated in bathing salt solutions at any ionic strength condition.

Multiscale Simulation of Protein Hydration Using the SWINGER Dynamical Clustering Algorithm.

To perform computationally efficient concurrent multiscale simulations of biological macromolecules in solution, where the all-atom (AT) models are coupled to supramolecular coarse-grained (SCG) solvent models, previous studies resorted to modified AT water models, such as the bundled-simple point charge (SPC) models, that use semiharmonic springs to restrict the relative movement of water molecules within a cluster. Those models can have a significant impact on the simulated biomolecules and can lead, for example, to a partial unfolding of a protein. In this work, we employ the recently developed alternative approach with a dynamical clustering algorithm, SWINGER, which enables a direct coupling of original unmodified AT and SCG water models. We perform an adaptive resolution molecular dynamics simulation of a Trp-Cage miniprotein in multiscale water, where the standard SPC water model is interfaced with the widely used MARTINI SCG model, and demonstrate that, compared to the corresponding full-blown AT simulations, the structural and dynamic properties of the solvated protein and surrounding solvent are well reproduced by our approach.

Adaptive resolution simulations of biomolecular systems.

In this review article, we discuss and analyze some recently developed hybrid atomistic-mesoscopic solvent models for multiscale biomolecular simulations. We focus on the biomolecular applications of the adaptive resolution scheme (AdResS), which allows solvent molecules to change their resolution back and forth between atomistic and coarse-grained representations according to their positions in the system. First, we discuss coupling of atomistic and coarse-grained models of salt solution using a 1-to-1 molecular mapping-i.e., one coarse-grained bead represents one water molecule-for development of a multiscale salt solution model. In order to make use of coarse-grained molecular models that are compatible with the MARTINI force field, one has to resort to a supramolecular mapping, in particular to a 4-to-1 mapping, where four water molecules are represented with one coarse-grained bead. To this end, bundled atomistic water models are employed, i.e., the relative movement of water molecules that are mapped to the same coarse-grained bead is restricted by employing harmonic springs. Supramolecular coupling has recently also been extended to polarizable coarse-grained water models with explicit charges. Since these coarse-grained models consist of several interaction sites, orientational degrees of freedom of the atomistic and coarse-grained representations are coupled via a harmonic energy penalty term. The latter aligns the dipole moments of both representations. The reviewed multiscale solvent models are ready to be used in biomolecular simulations, as illustrated in a few examples.

Adaptive resolution simulations coupling atomistic water to dissipative particle dynamics.

Multiscale methods are the most efficient way to address the interlinked spatiotemporal scales encountered in soft matter and molecular liquids. In the literature reported hybrid approaches span from quantum to atomistic, coarse-grained, and continuum length scales. In this article, we present the hybrid coupling of the molecular dynamics (MD) and dissipative particle dynamics (DPD) methods, bridging the micro- and mesoscopic descriptions. The interfacing is performed within the adaptive resolution scheme (AdResS), which is a linear momentum conserving coupling technique. Our methodology is hence suitable to simulate fluids on the micro/mesoscopic scale, where hydrodynamics plays an important role. The presented approach is showcased for water at ambient conditions. The supramolecular coupling is enabled by a recently developed clustering algorithm SWINGER that assembles, disassembles, and reassembles clusters as needed during the course of the simulation. This allows for a seamless coupling between standard atomistic MD and DPD models. The developed framework can be readily applied to various applications in the fields of materials and life sciences, e.g., simulations of phospholipids and polymer melts, or to study the red blood cells behavior in normal and disease states.

Order and interactions in DNA arrays: Multiscale molecular dynamics simulation.

While densely packed DNA arrays are known to exhibit hexagonal and orthorhombic local packings, the detailed mechanism governing the associated phase transition remains rather elusive. Furthermore, at high densities the atomistic resolution is paramount to properly account for fine details, encompassing the DNA molecular order, the contingent ordering of counterions and the induced molecular ordering of the bathing solvent, bringing together electrostatic, steric, thermal and direct hydrogen-bonding interactions, resulting in the observed osmotic equation of state. We perform a multiscale simulation of dense DNA arrays by enclosing a set of 16 atomistically resolved DNA molecules within a semi-permeable membrane, allowing the passage of water and salt ions, and thus mimicking the behavior of DNA arrays subjected to external osmotic stress in a bathing solution of monovalent salt and multivalent counterions. By varying the DNA density, local packing symmetry, and counterion type, we obtain osmotic equation of state together with the hexagonal-orthorhombic phase transition, and full structural characterization of the DNA subphase in terms of its positional and angular orientational fluctuations, counterion distributions, and the solvent local dielectric response profile with its order parameters that allow us to identify the hydration force as the primary interaction mechanism at high DNA densities.

Adaptive Resolution Simulation of Supramolecular Water: The Concurrent Making, Breaking, and Remaking of Water Bundles.

The adaptive resolution scheme (AdResS) is a multiscale molecular dynamics simulation approach that can concurrently couple atomistic (AT) and coarse-grained (CG) resolution regions, i.e., the molecules can freely adapt their resolution according to their current position in the system. Coupling to supramolecular CG models, where several molecules are represented as a single CG bead, is challenging, but it provides higher computational gains and connection to the established MARTINI CG force field. Difficulties that arise from such coupling have been so far bypassed with bundled AT water models, where additional harmonic bonds between oxygen atoms within a given supramolecular water bundle are introduced. While these models simplify the supramolecular coupling, they also cause in certain situations spurious artifacts, such as partial unfolding of biomolecules. In this work, we present a new clustering algorithm SWINGER that can concurrently make, break, and remake water bundles and in conjunction with the AdResS permits the use of original AT water models. We apply our approach to simulate a hybrid SPC/MARTINI water system and show that the essential properties of water are correctly reproduced with respect to the standard monoscale simulations. The developed hybrid water model can be used in biomolecular simulations, where a significant speed up can be obtained without compromising the accuracy of the AT water model.

Adaptive resolution simulation of a DNA molecule in salt solution.

We present a multiscale simulation of a DNA molecule in 1 M NaCl salt solution environment, employing the adaptive resolution simulation approach that allows the solvent molecules, i.e., water and ions, to change their resolution from atomistic to coarse-grained and vice versa adaptively on-the-fly. The region of high resolution moves together with the DNA center-of-mass so that the DNA itself is always modeled at high resolution. We show that our multiscale simulations yield a stable DNA-solution system, with statistical properties similar to those produced by the conventional all-atom molecular dynamics simulation. Special attention is given to the collective properties, such as the dielectric constant, as they provide a sensitive quality measure of our multiscale approach.

Adaptive resolution simulation of polarizable supramolecular coarse-grained water models.

Multiscale simulations methods, such as adaptive resolution scheme, are becoming increasingly popular due to their significant computational advantages with respect to conventional atomistic simulations. For these kind of simulations, it is essential to develop accurate multiscale water models that can be used to solvate biophysical systems of interest. Recently, a 4-to-1 mapping was used to couple the bundled-simple point charge water with the MARTINI model. Here, we extend the supramolecular mapping to coarse-grained models with explicit charges. In particular, the two tested models are the polarizable water and big multiple water models associated with the MARTINI force field. As corresponding coarse-grained representations consist of several interaction sites, we couple orientational degrees of freedom of the atomistic and coarse-grained representations via a harmonic energy penalty term. This additional energy term aligns the dipole moments of both representations. We test this coupling by studying the system under applied static external electric field. We show that our approach leads to the correct reproduction of the relevant structural and dynamical properties.

Adaptive resolution simulation of an atomistic protein in MARTINI water.

We present an adaptive resolution simulation of protein G in multiscale water. We couple atomistic water around the protein with mesoscopic water, where four water molecules are represented with one coarse-grained bead, farther away. We circumvent the difficulties that arise from coupling to the coarse-grained model via a 4-to-1 molecule coarse-grain mapping by using bundled water models, i.e., we restrict the relative movement of water molecules that are mapped to the same coarse-grained bead employing harmonic springs. The water molecules change their resolution from four molecules to one coarse-grained particle and vice versa adaptively on-the-fly. Having performed 15 ns long molecular dynamics simulations, we observe within our error bars no differences between structural (e.g., root-mean-squared deviation and fluctuations of backbone atoms, radius of gyration, the stability of native contacts and secondary structure, and the solvent accessible surface area) and dynamical properties of the protein in the adaptive resolution approach compared to the fully atomistically solvated model. Our multiscale model is compatible with the widely used MARTINI force field and will therefore significantly enhance the scope of biomolecular simulations.

Adaptive Resolution Simulation of MARTINI Solvents.

We present adaptive resolution molecular dynamics simulations of aqueous and apolar solvents using coarse-grained molecular models that are compatible with the MARTINI force field. As representatives of both classes of solvents we have chosen liquid water and butane, respectively, at ambient temperature. The solvent molecules change their resolution back and forth between the atomistic and coarse-grained representations according to their positions in the system. The difficulties that arise from coupling to a coarse-grained model with a multimolecule mapping, for example, 4-to-1 mapping in the case of the Simple Point Charge (SPC) and MARTINI water models, could be successfully circumvented by using bundled water models. We demonstrate that the presented multiscale approach faithfully reproduces the structural and dynamical properties computed by reference fully atomistic molecular dynamics simulations. Our approach is general and can be used with any atomistic force field to be linked with the MARTINI force field.