A systematic methodology for defining coarse-grained sites in large biomolecules

Evaluation of Essential Dynamics and Fixed-Length Coarse Graining for Multidomain Proteins

For multiscale modeling of biomolecules, reliable coarse-grained (CG) models can offer great potential to simulate larger temporal and spatial scales than traditional all-atom (AA) models. In this study, we explore the essential dynamics coarse graining (EDCG) and fixed-length coarse graining (FLCG) approaches for constructing highly coarse-grained models for multidomain proteins (MDPs), with 1 to 10 amino acid residues per CG site. In the studies of 13 MDPs, our data indicate that both EDCG and FLCG can preserve the protein dynamics of MDPs. FLCG, which restricts an equal number of residues in each CG site, represents an excellent approximation to EDCG and a straightforward approach for coarse-graining MDPs. Furthermore, FLCG is tested with a class B G-protein-coupled receptor protein, and the agreement with prior experiments suggests its general application to various MDPs in different environments or conditions. Finally, we demonstrate another application of FLCG through progressive backmapping, showcasing the ability to recover from lower-resolution CG models (6 residues/CG site) to higher-resolution ones (1 residue/CG site). These promising outcomes underscore the broad applicability of FLCG to construct highly or ultra-coarse-grained models of complex biomolecules for multiscale simulations.

Maximum Entropy Theory of Multiscale Coarse-Graining via Matching Thermodynamic Forces: Application to a Molecular Crystal (TATB)

The MSCG/FM (multiscale coarse-graining via force-matching) approach is an efficient supervised machine learning method to develop microscopically informed coarse-grained (CG) models. We present a theory based on the principle of maximum entropy (PME) enveloping the existing MSCG/FM approaches. This theory views the MSCG/FM method as a special case of matching the thermodynamic forces from the extended ensemble described by the set of thermodynamic (relevant) system coordinates. This set may include CG coordinates, the stress tensor, applied external fields, and so forth, and may be characterized by nonequilibrium conditions. Following the presentation of the theory, we discuss the consistent matching of both bonded and nonbonded interactions. The proposed PME formulation is used as a starting point to extend the MSCG/FM method to the constant strain ensemble, which together with the explicit matching of the bonded forces is better suited for coarse-graining anisotropic media at a submolecular resolution. The theory is demonstrated by performing the fine coarse-graining of crystalline 1,3,5-triamino-2,4,6-trinitrobenzene (TATB), a well-known insensitive molecular energetic material, which exhibits highly anisotropic mechanical properties.

Accelerated Lignocellulosic Molecule Adsorption Structure Determination

Here, we present a study combining Bayesian optimization structural inference with the machine learning interatomic potential Neural Equivariant Interatomic Potential (NequIP) to accelerate and enable the study of the adsorption of the conformationally flexible lignocellulosic molecules β-d-xylose and 1,4-β-d-xylotetraose on a copper surface. The number of structure evaluations needed to map out the relevant potential energy surfaces are reduced by Bayesian optimization, while NequIP minimizes the time spent on each evaluation, ultimately resulting in cost-efficient and reliable sampling of large systems and configurational spaces. Although the applicability of Bayesian optimization for the conformational analysis of the more flexible xylotetraose molecule is restricted by the sample complexity bottleneck, the latter can be effectively bypassed with external conformer search tools, such as the Conformer-Rotamer Ensemble Sampling Tool, facilitating the subsequent lower-dimensional global minimum adsorption structure determination. Finally, we demonstrate the applicability of the described approach to find adsorption structures practically equivalent to the density functional theory counterparts at a fraction of the computational cost.

Formalizing Coarse-Grained Representations of Anisotropic Interactions at Multimeric Protein Interfaces Using Virtual Sites

Molecular simulations of biomacromolecules that assemble into multimeric complexes remain a challenge due to computationally inaccessible length and time scales. Low-resolution and implicit-solvent coarse-grained modeling approaches using traditional nonbonded interactions (both pairwise and spherically isotropic) have been able to partially address this gap. However, these models may fail to capture the complex anisotropic interactions present at macromolecular interfaces unless higher-order interaction potentials are incorporated at the expense of the computational cost. In this work, we introduce an alternate and systematic approach to represent directional interactions at protein-protein interfaces by using virtual sites restricted to pairwise interactions. We show that virtual site interaction parameters can be optimized within a relative entropy minimization framework by using only information from known statistics between coarse-grained sites. We compare our virtual site models to traditional coarse-grained models using two case studies of multimeric protein assemblies and find that the virtual site models predict pairwise correlations with higher fidelity and, more importantly, assembly behavior that is morphologically consistent with experiments. Our study underscores the importance of anisotropic interaction representations and paves the way for more accurate yet computationally efficient coarse-grained simulations of macromolecular assembly in future research.

Surveying the energy landscape of coarse-grained mappings

Simulations of soft materials often adopt low-resolution coarse-grained (CG) models. However, the CG representation is not unique and its impact upon simulated properties is poorly understood. In this work, we investigate the space of CG representations for ubiquitin, which is a typical globular protein with 72 amino acids. We employ Monte Carlo methods to ergodically sample this space and to characterize its landscape. By adopting the Gaussian network model as an analytically tractable atomistic model for equilibrium fluctuations, we exactly assess the intrinsic quality of each CG representation without introducing any approximations in sampling configurations or in modeling interactions. We focus on two metrics, the spectral quality and the information content, that quantify the extent to which the CG representation preserves low-frequency, large-amplitude motions and configurational information, respectively. The spectral quality and information content are weakly correlated among high-resolution representations but become strongly anticorrelated among low-resolution representations. Representations with maximal spectral quality appear consistent with physical intuition, while low-resolution representations with maximal information content do not. Interestingly, quenching studies indicate that the energy landscape of mapping space is very smooth and highly connected. Moreover, our study suggests a critical resolution below which a "phase transition" qualitatively distinguishes good and bad representations.

HIV-1 capsid shape, orientation, and entropic elasticity regulate translocation into the nuclear pore complex

Nuclear import and uncoating of the viral capsid are critical steps in the HIV-1 life cycle that serve to transport and release genomic material into the nucleus. Viral core import involves translocating the HIV-1 capsid at the nuclear pore complex (NPC). Notably, the central channel of the NPC appears to often accommodate and allow passage of intact HIV-1 capsid, though mechanistic details of the process remain to be fully understood. Here, we investigate the molecular interactions that operate in concert between the HIV-1 capsid and the NPC that regulate capsid translocation through the central channel. To this end, we develop a "bottom-up" coarse-grained (CG) model of the human NPC from recently released cryo-electron tomography structure and then construct composite membrane-embedded CG NPC models. We find that successful translocation from the cytoplasmic side to the NPC central channel is contingent on the compatibility of the capsid morphology and channel dimension and the proper orientation of the capsid approach to the channel from the cytoplasmic side. The translocation dynamics is driven by maximizing the contacts between phenylalanine-glycine nucleoporins at the central channel and the capsid. For the docked intact capsids, structural analysis reveals correlated striated patterns of lattice disorder likely related to the intrinsic capsid elasticity. Uncondensed genomic material inside the docked capsid augments the overall lattice disorder of the capsid. Our results suggest that the intrinsic "elasticity" can also aid the capsid to adapt to the stress and remain structurally intact during translocation.

Conformational transitions of the HIV-1 Gag polyprotein upon multimerization and gRNA binding

During the HIV-1 assembly process, the Gag polyprotein multimerizes at the producer cell plasma membrane, resulting in the formation of spherical immature virus particles. Gag-genomic RNA (gRNA) interactions play a crucial role in the multimerization process, which is yet to be fully understood. We performed large-scale all-atom molecular dynamics simulations of membrane-bound full-length Gag dimer, hexamer, and 18-mer. The inter-domain dynamic correlation of Gag, quantified by the heterogeneous elastic network model applied to the simulated trajectories, is observed to be altered by implicit gRNA binding, as well as the multimerization state of the Gag. The lateral dynamics of our simulated membrane-bound Gag proteins, with and without gRNA binding, agree with prior experimental data and help to validate our simulation models and methods. The gRNA binding is observed to affect mainly the SP1 domain of the 18-mer and the matrix-capsid linker domain of the hexamer. In the absence of gRNA binding, the independent dynamical motion of the nucleocapsid domain results in a collapsed state of the dimeric Gag. Unlike stable SP1 helices in the six-helix bundle, without IP6 binding, the SP1 domain undergoes a spontaneous helix-to-coil transition in the dimeric Gag. Together, our findings reveal conformational switches of Gag at different stages of the multimerization process and predict that the gRNA binding reinforces an efficient binding surface of Gag for multimerization, and also regulates the dynamic organization of the local membrane region itself.

K-Means Clustering Coarse-Graining (KMC-CG): A Next Generation Methodology for Determining Optimal Coarse-Grained Mappings of Large Biomolecules

Coarse-grained (CG) molecular dynamics (MD) has become a method of choice for simulating various large scale biomolecular processes; therefore, the systematic definition of the CG mappings for biomolecules remains an important topic. Appropriate CG mappings can significantly enhance the representability of a CG model and improve its ability to capture critical features of large biomolecules. In this work, we present a systematic and more generalized method called K-means clustering coarse-graining (KMC-CG), which builds on the earlier approach of essential dynamics coarse-graining (ED-CG). KMC-CG removes the sequence-dependent constraints of ED-CG, allowing it to explore a more extensive space and thus enabling the discovery of more physically optimal CG mappings. Furthermore, the implementation of the K-means clustering algorithm can variationally optimize the CG mapping with efficiency and stability. This new method is tested in three cases: ATP-bound G-actin, the HIV-1 CA pentamer, and the Arp2/3 complex. In these examples, the CG models generated by KMC-CG are seen to better capture the structural, dynamic, and functional domains. KMC-CG therefore provides a robust and consistent approach to generating CG models of large biomolecules that can then be more accurately parametrized by either bottom-up or top-down CG force fields.

OpenMSCG: A Software Tool for Bottom-Up Coarse-Graining

The "bottom-up" approach to coarse-graining, for building accurate and efficient computational models to simulate large-scale and complex phenomena and processes, is an important approach in computational chemistry, biophysics, and materials science. As one example, the Multiscale Coarse-Graining (MS-CG) approach to developing CG models can be rigorously derived using statistical mechanics applied to fine-grained, i.e., all-atom simulation data for a given system. Under a number of circumstances, a systematic procedure, such as MS-CG modeling, is particularly valuable. Here, we present the development of the OpenMSCG software, a modularized open-source software that provides a collection of successful and widely applied bottom-up CG methods, including Boltzmann Inversion (BI), Force-Matching (FM), Ultra-Coarse-Graining (UCG), Relative Entropy Minimization (REM), Essential Dynamics Coarse-Graining (EDCG), and Heterogeneous Elastic Network Modeling (HeteroENM). OpenMSCG is a high-performance and comprehensive toolset that can be used to derive CG models from large-scale fine-grained simulation data in file formats from common molecular dynamics (MD) software packages, such as GROMACS, LAMMPS, and NAMD. OpenMSCG is modularized in the Python programming framework, which allows users to create and customize modeling "recipes" for reproducible results, thus greatly improving the reliability, reproducibility, and sharing of bottom-up CG models and their applications.

Neural potentials of proteins extrapolate beyond training data

We evaluate neural network (NN) coarse-grained (CG) force fields compared to traditional CG molecular mechanics force fields. We conclude that NN force fields are able to extrapolate and sample from unseen regions of the free energy surface when trained with limited data. Our results come from 88 NN force fields trained on different combinations of clustered free energy surfaces from four protein mapped trajectories. We used a statistical measure named total variation similarity to assess the agreement between reference free energy surfaces from mapped atomistic simulations and CG simulations from trained NN force fields. Our conclusions support the hypothesis that NN CG force fields trained with samples from one region of the proteins' free energy surface can, indeed, extrapolate to unseen regions. Additionally, the force matching error was found to only be weakly correlated with a force field's ability to reconstruct the correct free energy surface.

Developing Hybrid All-Atom and Ultra-Coarse-Grained Models to Investigate Taxol-Binding and Dynein Interactions on Microtubules

Simulating the conformations and functions of biological macromolecules by using all-atom (AA) models is a challenging task due to expensive computational costs. One possible strategy to solve this problem is to develop hybrid all-atom and ultra-coarse-grained (AA/UCG) models of the biological macromolecules. In the AA/UCG scheme, the interest regions are described by AA models, while the other regions are described in the UCG representation. In this study, we develop the hybrid AA/UCG models and apply them to investigate the conformational changes of microtubule-bound tubulins. The simulation results of the hybrid models elucidated the mechanism of why the taxol molecules selectively bound microtubules but not tubulin dimers. In addition, we also explore the interactions of the microtubules and dyneins. Our study shows that the hybrid AA/UCG model has great application potential in studying the function of complex biological systems.

Conformational transitions of the HIV-1 Gag polyprotein upon multimerization and gRNA binding

During the HIV-1 assembly process, the Gag polyprotein multimerizes at the producer cell plasma membrane, resulting in the formation of spherical immature virus particles. Gag-gRNA interactions play a crucial role in the multimerization process, which is yet to be fully understood. We have performed large-scale all-atom molecular dynamics simulations of membrane-bound full-length Gag dimer, hexamer, and 18-mer. The inter-domain dynamic correlation of Gag, quantified by the heterogeneous elastic network model (hENM) applied to the simulated trajectories, is observed to be altered by implicit gRNA binding, as well as the multimerization state of the Gag. The lateral dynamics of our simulated membrane-bound Gag proteins, with and without gRNA binding, agree with prior experimental data and help to validate our simulation models and methods. The gRNA binding is observed to impact mainly the SP1 domain of the 18-mer and the MA-CA linker domain of the hexamer. In the absence of gRNA binding, the independent dynamical motion of the NC domain results in a collapsed state of the dimeric Gag. Unlike stable SP1 helices in the six-helix bundle, without IP6 binding, the SP1 domain undergoes a spontaneous helix-to-coil transition in the dimeric Gag. Together, our findings reveal conformational switches of Gag at different stages of the multimerization process and predict that the gRNA binding reinforces an efficient binding surface of Gag for multimerization, as well as regulates the dynamic organization of the local membrane region itself.

Significance

Gag(Pr 55 Gag ) polyprotein orchestrates many essential events in HIV-1 assembly, including packaging of the genomic RNA (gRNA) in the immature virion. Although various experimental techniques, such as cryo-ET, X-ray, and NMR, have revealed structural properties of individual domains in the immature Gag clusters, structural and biophysical characterization of a full-length Gag molecule remains a challenge for existing experimental techniques. Using atomistic molecular dynamics simulations of the different model systems of Gag polyprotein, we present here a detailed structural characterization of Gag molecules in different multimerization states and interrogate the synergy between Gag-Gag, Gag-membrane, and Gag-gRNA interactions during the viral assembly process.

Utilizing Machine Learning to Greatly Expand the Range and Accuracy of Bottom-Up Coarse-Grained Models through Virtual Particles

Coarse-grained (CG) models parametrized using atomistic reference data, i.e., "bottom up" CG models, have proven useful in the study of biomolecules and other soft matter. However, the construction of highly accurate, low resolution CG models of biomolecules remains challenging. We demonstrate in this work how virtual particles, CG sites with no atomistic correspondence, can be incorporated into CG models within the context of relative entropy minimization (REM) as latent variables. The methodology presented, variational derivative relative entropy minimization (VD-REM), enables optimization of virtual particle interactions through a gradient descent algorithm aided by machine learning. We apply this methodology to the challenging case of a solvent-free CG model of a 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) lipid bilayer and demonstrate that introduction of virtual particles captures solvent-mediated behavior and higher-order correlations which REM alone cannot capture in a more standard CG model based only on the mapping of collections of atoms to the CG sites.

Automated Parameterization of Coarse-Grained Polyethylenimine under a Martini Framework

As a versatile polymer in many applications, synthesized polyethylenimine (PEI) is polydisperse with diverse branched structures that attain pH-dependent protonation states. Understanding the structure-function relationship of PEI is necessary for enhancing its efficacy in various applications. Coarse-grained (CG) simulations can be performed at length and time scales directly comparable with experimental data while maintaining the molecular perspective. However, manually developing CG forcefields for complex PEI structures is time-consuming and prone to human errors. This article presents a fully automated algorithm that can coarse-grain any branched architecture of PEI from its all-atom (AA) simulation trajectories and topology. The algorithm is demonstrated by coarse-graining a branched 2 kDa PEI, which can replicate the AA diffusion coefficient, radius of gyration, and end-to-end distance of the longest linear chain. Commercially available 25 and 2 kDa Millipore-Sigma PEIs are used for experimental validation. Specifically, branched PEI architectures are proposed, coarse-grained using the automated algorithm, and then simulated at different mass concentrations. The CG PEIs can reproduce existing experimental data on PEI's diffusion coefficient and Stokes-Einstein radius at infinite dilution as well as its intrinsic viscosity. This suggests a strategy where probable chemical structures of synthetic PEIs can be inferred computationally using the developed algorithm. The coarse-graining methodology presented here can also be extended to other polymers.

Dynamics of upstream ESCRT organization at the HIV-1 budding site

In the late stages of the HIV-1 life cycle, membrane localization and self-assembly of Gag polyproteins induce membrane deformation and budding. Release of the virion requires direct interaction between immature Gag lattice and upstream ESCRT machinery at the viral budding site, followed by assembly of downstream ESCRT-III factors, culminating in membrane scission. However, molecular details of upstream ESCRT assembly dynamics at the viral budding site remain unclear. In this work, using coarse-grained (CG) molecular dynamics (MD) simulations, we investigated the interactions between Gag, ESCRT-I, ESCRT-II, and membrane to delineate the dynamical mechanisms by which upstream ESCRTs assemble templated by late-stage immature Gag lattice. We first systematically derived "bottom-up" CG molecular models and interactions of upstream ESCRT proteins from experimental structural data and extensive all-atom MD simulations. Using these molecular models, we performed CG MD simulations of ESCRT-I oligomerization and ESCRT-I/II supercomplex formation at the neck of the budding virion. Our simulations demonstrate that ESCRT-I can effectively oligomerize to higher-order complexes templated by the immature Gag lattice both in the absence of ESCRT-II and when multiple copies of ESCRT-II are localized at the bud neck. The ESCRT-I/II supercomplexes formed in our simulations exhibit predominantly columnar structures, which has important implications for the nucleation pathway of downstream ESCRT-III polymers. Importantly, ESCRT-I/II supercomplexes bound to Gag initiate membrane neck constriction by pulling the inner edge of the bud neck closer to the ESCRT-I headpiece ring. Our findings serve to elucidate a network of interactions between upstream ESCRT machinery, immature Gag lattice, and membrane neck that regulate protein assembly dynamics at the HIV-1 budding site.

Statistically Optimal Force Aggregation for Coarse-Graining Molecular Dynamics

Machine-learned coarse-grained (CG) models have the potential for simulating large molecular complexes beyond what is possible with atomistic molecular dynamics. However, training accurate CG models remains a challenge. A widely used methodology for learning bottom-up CG force fields maps forces from all-atom molecular dynamics to the CG representation and matches them with a CG force field on average. We show that there is flexibility in how to map all-atom forces to the CG representation and that the most commonly used mapping methods are statistically inefficient and potentially even incorrect in the presence of constraints in the all-atom simulation. We define an optimization statement for force mappings and demonstrate that substantially improved CG force fields can be learned from the same simulation data when using optimized force maps. The method is demonstrated on the miniproteins chignolin and tryptophan cage and published as open-source code.

Ensuring thermodynamic consistency with invertible coarse-graining

Coarse-grained models are a core computational tool in theoretical chemistry and biophysics. A judicious choice of a coarse-grained model can yield physical insights by isolating the essential degrees of freedom that dictate the thermodynamic properties of a complex, condensed-phase system. The reduced complexity of the model typically leads to lower computational costs and more efficient sampling compared with atomistic models. Designing "good" coarse-grained models is an art. Generally, the mapping from fine-grained configurations to coarse-grained configurations itself is not optimized in any way; instead, the energy function associated with the mapped configurations is. In this work, we explore the consequences of optimizing the coarse-grained representation alongside its potential energy function. We use a graph machine learning framework to embed atomic configurations into a low-dimensional space to produce efficient representations of the original molecular system. Because the representation we obtain is no longer directly interpretable as a real-space representation of the atomic coordinates, we also introduce an inversion process and an associated thermodynamic consistency relation that allows us to rigorously sample fine-grained configurations conditioned on the coarse-grained sampling. We show that this technique is robust, recovering the first two moments of the distribution of several observables in proteins such as chignolin and alanine dipeptide.

Integrating Machine Learning in the Coarse-Grained Molecular Simulation of Polymers

Machine learning (ML) is having an increasing impact on the physical sciences, engineering, and technology and its integration into molecular simulation frameworks holds great potential to expand their scope of applicability to complex materials and facilitate fundamental knowledge and reliable property predictions, contributing to the development of efficient materials design routes. The application of ML in materials informatics in general, and polymer informatics in particular, has led to interesting results, however great untapped potential lies in the integration of ML techniques into the multiscale molecular simulation methods for the study of macromolecular systems, specifically in the context of Coarse Grained (CG) simulations. In this Perspective, we aim at presenting the pioneering recent research efforts in this direction and discussing how these new ML-based techniques can contribute to critical aspects of the development of multiscale molecular simulation methods for bulk complex chemical systems, especially polymers. Prerequisites for the implementation of such ML-integrated methods and open challenges that need to be met toward the development of general systematic ML-based coarse graining schemes for polymers are discussed.

Bottom-up Coarse-Graining: Principles and Perspectives

Large-scale computational molecular models provide scientists a means to investigate the effect of microscopic details on emergent mesoscopic behavior. Elucidating the relationship between variations on the molecular scale and macroscopic observable properties facilitates an understanding of the molecular interactions driving the properties of real world materials and complex systems (e.g., those found in biology, chemistry, and materials science). As a result, discovering an explicit, systematic connection between microscopic nature and emergent mesoscopic behavior is a fundamental goal for this type of investigation. The molecular forces critical to driving the behavior of complex heterogeneous systems are often unclear. More problematically, simulations of representative model systems are often prohibitively expensive from both spatial and temporal perspectives, impeding straightforward investigations over possible hypotheses characterizing molecular behavior. While the reduction in resolution of a study, such as moving from an atomistic simulation to that of the resolution of large coarse-grained (CG) groups of atoms, can partially ameliorate the cost of individual simulations, the relationship between the proposed microscopic details and this intermediate resolution is nontrivial and presents new obstacles to study. Small portions of these complex systems can be realistically simulated. Alone, these smaller simulations likely do not provide insight into collectively emergent behavior. However, by proposing that the driving forces in both smaller and larger systems (containing many related copies of the smaller system) have an explicit connection, systematic bottom-up CG techniques can be used to transfer CG hypotheses discovered using a smaller scale system to a larger system of primary interest. The proposed connection between different CG systems is prescribed by (i) the CG representation (mapping) and (ii) the functional form and parameters used to represent the CG energetics, which approximate potentials of mean force (PMFs). As a result, the design of CG methods that facilitate a variety of physically relevant representations, approximations, and force fields is critical to moving the frontier of systematic CG forward. Crucially, the proposed connection between the system used for parametrization and the system of interest is orthogonal to the optimization used to approximate the potential of mean force present in all systematic CG methods. The empirical efficacy of machine learning techniques on a variety of tasks provides strong motivation to consider these approaches for approximating the PMF and analyzing these approximations.

Inositol Hexakisphosphate (IP6) Accelerates Immature HIV-1 Gag Protein Assembly toward Kinetically Trapped Morphologies

During the late stages of the HIV-1 lifecycle, immature virions are produced by the concerted activity of Gag polyproteins, primarily mediated by the capsid (CA) and spacer peptide 1 (SP1) domains, which assemble into a spherical lattice, package viral genomic RNA, and deform the plasma membrane. Recently, inositol hexakisphosphate (IP6) has been identified as an essential assembly cofactor that efficiently produces both immature virions in vivo and immature virus-like particles in vitro. To date, however, several distinct mechanistic roles for IP6 have been proposed on the basis of independent functional, structural, and kinetic studies. In this work, we investigate the molecular influence of IP6 on the structural outcomes and dynamics of CA/SP1 assembly using coarse-grained (CG) molecular dynamics (MD) simulations and free energy calculations. Here, we derive a bottom-up, low-resolution, and implicit-solvent CG model of CA/SP1 and IP6, and simulate their assembly under conditions that emulate both in vitro and in vivo systems. Our analysis identifies IP6 as an assembly accelerant that promotes curvature generation and fissure-like defects throughout the lattice. Our findings suggest that IP6 induces kinetically trapped immature morphologies, which may be physiologically important for later stages of viral morphogenesis and potentially useful for virus-like particle technologies.

Current State and Perspectives of Simulation and Modeling of Aliphatic Isocyanates and Polyisocyanates

Aliphatic isocyanates and polyisocyanates are central molecules in the fabrication of polyurethanes, coatings, and adhesives and, due to their excellent mechanical and stability properties, are continuously investigated in advanced applications; however, despite the growing interest in isocyanate-based systems, atomistic simulations on them have been limited by the lack of accurate parametrizations for these molecular species. In this review, we will first provide an overview of current research on isocyanate systems to highlight their most promising applications, especially in fields far from their typical usage, and to justify the need for further modeling works. Next, we will discuss the state of their modeling, from first-principle studies to atomistic molecular dynamics simulations and coarse-grained approaches, highlighting the recent advances in atomistic modeling. Finally, the most promising lines of research in the modeling of isocyanates are discussed in light of the possibilities opened by novel approaches, such as machine learning.

Cooperative multivalent receptor binding promotes exposure of the SARS-CoV-2 fusion machinery core

The molecular events that permit the spike glycoprotein of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) to bind and enter cells are important to understand for both fundamental and therapeutic reasons. Spike proteins consist of S1 and S2 domains, which recognize angiotensin-converting enzyme 2 (ACE2) receptors and contain the viral fusion machinery, respectively. Ostensibly, the binding of spike trimers to ACE2 receptors promotes dissociation of the S1 domains and exposure of the fusion machinery, although the molecular details of this process have yet to be observed. We report the development of bottom-up coarse-grained (CG) models consistent with cryo-electron tomography data, and the use of CG molecular dynamics simulations to investigate viral binding and S2 core exposure. We show that spike trimers cooperatively bind to multiple ACE2 dimers at virion-cell interfaces in a manner distinct from binding between soluble proteins, which processively induces S1 dissociation. We also simulate possible variant behavior using perturbed CG models, and find that ACE2-induced S1 dissociation is primarily sensitive to conformational state populations and the extent of S1/S2 cleavage, rather than ACE2 binding affinity. These simulations reveal an important concerted interaction between spike trimers and ACE2 dimers that primes the virus for membrane fusion and entry.

Integrin-based mechanosensing through conformational deformation

Conversion of integrins from low to high affinity states, termed activation, is important in biological processes, including immunity, hemostasis, angiogenesis, and embryonic development. Integrin activation is regulated by large-scale conformational transitions from closed, low affinity states to open, high affinity states. Although it has been suggested that substrate stiffness shifts the conformational equilibrium of integrin and governs its unbinding, here, we address the role of integrin conformational activation in cellular mechanosensing. Comparison of wild-type versus activating mutants of integrin αVβ3 show that activating mutants shift cell spreading, focal adhesion kinase activation, traction stress, and force on talin toward high stiffness values at lower stiffness. Although all activated integrin mutants showed equivalent binding affinity for soluble ligands, the β3 S243E mutant showed the strongest shift in mechanical responses. To understand this behavior, we used coarse-grained computational models derived from molecular level information. The models predicted that wild-type integrin αVβ3 displaces under force and that activating mutations shift the required force toward lower values, with S243E showing the strongest effect. Cellular stiffness sensing thus correlates with computed effects of force on integrin conformation. Together, these data identify a role for force-induced integrin conformational deformation in cellular mechanosensing.

From System Modeling to System Analysis: The Impact of Resolution Level and Resolution Distribution in the Computer-Aided Investigation of Biomolecules

The ever increasing computer power, together with the improved accuracy of atomistic force fields, enables researchers to investigate biological systems at the molecular level with remarkable detail. However, the relevant length and time scales of many processes of interest are still hardly within reach even for state-of-the-art hardware, thus leaving important questions often unanswered. The computer-aided investigation of many biological physics problems thus largely benefits from the usage of coarse-grained models, that is, simplified representations of a molecule at a level of resolution that is lower than atomistic. A plethora of coarse-grained models have been developed, which differ most notably in their granularity; this latter aspect determines one of the crucial open issues in the field, i.e. the identification of an optimal degree of coarsening, which enables the greatest simplification at the expenses of the smallest information loss. In this review, we present the problem of coarse-grained modeling in biophysics from the viewpoint of system representation and information content. In particular, we discuss two distinct yet complementary aspects of protein modeling: on the one hand, the relationship between the resolution of a model and its capacity of accurately reproducing the properties of interest; on the other hand, the possibility of employing a lower resolution description of a detailed model to extract simple, useful, and intelligible information from the latter.

Cooperative multivalent receptor binding promotes exposure of the SARS-CoV-2 fusion machinery core

The molecular events that permit the spike glycoprotein of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) to bind, fuse, and enter cells are important to understand for both fundamental and therapeutic reasons. Spike proteins consist of S1 and S2 domains, which recognize angiotensin-converting enzyme 2 (ACE2) receptors and contain the viral fusion machinery, respectively. Ostensibly, the binding of spike trimers to ACE2 receptors promotes the preparation of the fusion machinery by dissociation of the S1 domains. We report the development of bottom-up coarse-grained (CG) models validated with cryo-electron tomography (cryo-ET) data, and the use of CG molecular dynamics simulations to investigate the dynamical mechanisms involved in viral binding and exposure of the S2 trimeric core. We show that spike trimers cooperatively bind to multiple ACE2 dimers at virion-cell interfaces. The multivalent interaction cyclically and processively induces S1 dissociation, thereby exposing the S2 core containing the fusion machinery. Our simulations thus reveal an important concerted interaction between spike trimers and ACE2 dimers that primes the virus for membrane fusion and entry.

Bottom-Up Coarse-Grained Modeling of DNA

Recent advances in methodology enable effective coarse-grained modeling of deoxyribonucleic acid (DNA) based on underlying atomistic force field simulations. The so-called bottom-up coarse-graining practice separates fast and slow dynamic processes in molecular systems by averaging out fast degrees of freedom represented by the underlying fine-grained model. The resulting effective potential of interaction includes the contribution from fast degrees of freedom effectively in the form of potential of mean force. The pair-wise additive potential is usually adopted to construct the coarse-grained Hamiltonian for its efficiency in a computer simulation. In this review, we present a few well-developed bottom-up coarse-graining methods, discussing their application in modeling DNA properties such as DNA flexibility (persistence length), conformation, "melting," and DNA condensation.

A multiscale coarse-grained model of the SARS-CoV-2 virion

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is the causative agent of the COVID-19 pandemic. Computer simulations of complete viral particles can provide theoretical insights into large-scale viral processes including assembly, budding, egress, entry, and fusion. Detailed atomistic simulations are constrained to shorter timescales and require billion-atom simulations for these processes. Here, we report the current status and ongoing development of a largely "bottom-up" coarse-grained (CG) model of the SARS-CoV-2 virion. Data from a combination of cryo-electron microscopy (cryo-EM), x-ray crystallography, and computational predictions were used to build molecular models of structural SARS-CoV-2 proteins, which were then assembled into a complete virion model. We describe how CG molecular interactions can be derived from all-atom simulations, how viral behavior difficult to capture in atomistic simulations can be incorporated into the CG models, and how the CG models can be iteratively improved as new data become publicly available. Our initial CG model and the detailed methods presented are intended to serve as a resource for researchers working on COVID-19 who are interested in performing multiscale simulations of the SARS-CoV-2 virion.

Coarse-Grained Simulation of Mechanical Properties of Single Microtubules With Micrometer Length

Microtubules are one of the most important components in the cytoskeleton and play a vital role in maintaining the shape and function of cells. Because single microtubules are some micrometers long, it is difficult to simulate such a large system using an all-atom model. In this work, we use the newly developed convolutional and K-means coarse-graining (CK-CG) method to establish an ultra-coarse-grained (UCG) model of a single microtubule, on the basis of the low electron microscopy density data of microtubules. We discuss the rationale of the micro-coarse-grained microtubule models of different resolutions and explore microtubule models up to 12-micron length. We use the devised microtubule model to quantify mechanical properties of microtubules of different lengths. Our model allows mesoscopic simulations of micrometer-level biomaterials and can be further used to study important biological processes related to microtubule function.

Machine Learning of Coarse-Grained Models for Organic Molecules and Polymers: Progress, Opportunities, and Challenges

Machine learning (ML) has emerged as one of the most powerful tools transforming all areas of science and engineering. The nature of molecular dynamics (MD) simulations, complex and time-consuming calculations, makes them particularly suitable for ML research. This review article focuses on recent advancements in developing efficient and accurate coarse-grained (CG) models using various ML methods, in terms of regulating the coarse-graining process, constructing adequate descriptors/features, generating representative training data sets, and optimization of the loss function. Two classes of the CG models are introduced: bottom-up and top-down CG methods. To illustrate these methods and demonstrate the open methodological questions, we survey several important principles in constructing CG models and how these are incorporated into ML methods and improved with specific learning techniques. Finally, we discuss some key aspects of developing machine-learned CG models with high accuracy and efficiency. Besides, we describe how these aspects are tackled in state-of-the-art methods and which remain to be addressed in the near future. We expect that these machine-learned CG models can address thermodynamic consistent, transferable, and representative issues in classical CG models.

An Information-Theory-Based Approach for Optimal Model Reduction of Biomolecules

In theoretical modeling of a physical system, a crucial step consists of the identification of those degrees of freedom that enable a synthetic yet informative representation of it. While in some cases this selection can be carried out on the basis of intuition and experience, straightforward discrimination of the important features from the negligible ones is difficult for many complex systems, most notably heteropolymers and large biomolecules. We here present a thermodynamics-based theoretical framework to gauge the effectiveness of a given simplified representation by measuring its information content. We employ this method to identify those reduced descriptions of proteins, in terms of a subset of their atoms, that retain the largest amount of information from the original model; we show that these highly informative representations share common features that are intrinsically related to the biological properties of the proteins under examination, thereby establishing a bridge between protein structure, energetics, and function.

A Multiscale Coarse-grained Model of the SARS-CoV-2 Virion

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is the causative agent of the COVID-19 pandemic. Computer simulations of complete viral particles can provide theoretical insights into large-scale viral processes including assembly, budding, egress, entry, and fusion. Detailed atomistic simulations, however, are constrained to shorter timescales and require billion-atom simulations for these processes. Here, we report the current status and on-going development of a largely "bottom-up" coarse-grained (CG) model of the SARS-CoV-2 virion. Structural data from a combination of cryo-electron microscopy (cryo-EM), x-ray crystallography, and computational predictions were used to build molecular models of structural SARS-CoV-2 proteins, which were then assembled into a complete virion model. We describe how CG molecular interactions can be derived from all-atom simulations, how viral behavior difficult to capture in atomistic simulations can be incorporated into the CG models, and how the CG models can be iteratively improved as new data becomes publicly available. Our initial CG model and the detailed methods presented are intended to serve as a resource for researchers working on COVID-19 who are interested in performing multiscale simulations of the SARS-CoV-2 virion.

Significance Statement

This study reports the construction of a molecular model for the SARS-CoV-2 virion and details our multiscale approach towards model refinement. The resulting model and methods can be applied to and enable the simulation of SARS-CoV-2 virions.

Exploring the landscape of model representations

The success of any physical model critically depends upon adopting an appropriate representation for the phenomenon of interest. Unfortunately, it remains generally challenging to identify the essential degrees of freedom or, equivalently, the proper order parameters for describing complex phenomena. Here we develop a statistical physics framework for exploring and quantitatively characterizing the space of order parameters for representing physical systems. Specifically, we examine the space of low-resolution representations that correspond to particle-based coarse-grained (CG) models for a simple microscopic model of protein fluctuations. We employ Monte Carlo (MC) methods to sample this space and determine the density of states for CG representations as a function of their ability to preserve the configurational information, I, and large-scale fluctuations, Q, of the microscopic model. These two metrics are uncorrelated in high-resolution representations but become anticorrelated at lower resolutions. Moreover, our MC simulations suggest an emergent length scale for coarse-graining proteins, as well as a qualitative distinction between good and bad representations of proteins. Finally, we relate our work to recent approaches for clustering graphs and detecting communities in networks.

Double-Well Ultra-Coarse-Grained Model to Describe Protein Conformational Transitions

The double-well model is usually used to describe the conformational transition between two states of a protein. Since conformational changes usually occur within a relatively large time scale, coarse-grained models are often used to accelerate the dynamic process due to their inexpensive computational cost. In this work, we develop a double-well ultra-coarse-grained (DW-UCG) model to describe the conformational transitions of the adenylate kinase, glutamine-binding protein, and lactoferrin. The coarse-grained simulation results show that the DW-UCG model of adenylate kinase captures the crucial intermediate states in the LID-closing and NMP-closing pathways, reflecting the key secondary structural changes in the conformational transition. A comparison of the different DW-UCG models of adenylate kinase indicates that an appropriate choice of bead resolution could generate the free energy landscape that is comparable to that from the residue-based model. The coarse-grained simulations for the glutamine-binding protein and lactoferrin also demonstrate that the DW-UCG model is valid in reproducing the correct two-state behavior for their functional study, which indicates the potential application of the DW-UCG model in investigating the mechanism of conformational changes of large proteins.

Graph neural network based coarse-grained mapping prediction

The selection of coarse-grained (CG) mapping operators is a critical step for CG molecular dynamics (MD) simulation. It is still an open question about what is optimal for this choice and there is a need for theory. The current state-of-the art method is mapping operators manually selected by experts. In this work, we demonstrate an automated approach by viewing this problem as supervised learning where we seek to reproduce the mapping operators produced by experts. We present a graph neural network based CG mapping predictor called Deep Supervised Graph Partitioning Model (DSGPM) that treats mapping operators as a graph segmentation problem. DSGPM is trained on a novel dataset, Human-annotated Mappings (HAM), consisting of 1180 molecules with expert annotated mapping operators. HAM can be used to facilitate further research in this area. Our model uses a novel metric learning objective to produce high-quality atomic features that are used in spectral clustering. The results show that the DSGPM outperforms state-of-the-art methods in the field of graph segmentation. Finally, we find that predicted CG mapping operators indeed result in good CG MD models when used in simulation.

We propose a scalable graph neural network-based method for automating coarse-grained mapping prediction for molecules.

Ligand-protein interactions in lysozyme investigated through a dual-resolution model

A fully atomistic (AT) modeling of biological macromolecules at relevant length- and time-scales is often cumbersome or not even desirable, both in terms of computational effort required and a posteriori analysis. This difficulty can be overcome with the use of multiresolution models, in which different regions of the same system are concurrently described at different levels of detail. In enzymes, computationally expensive AT detail is crucial in the modeling of the active site in order to capture, for example, the chemically subtle process of ligand binding. In contrast, important yet more collective properties of the remainder of the protein can be reproduced with a coarser description. In the present work, we demonstrate the effectiveness of this approach through the calculation of the binding free energy of hen egg white lysozyme with the inhibitor di-N-acetylchitotriose. Particular attention is payed to the impact of the mapping, that is, the selection of AT and coarse-grained residues, on the binding free energy. It is shown that, in spite of small variations of the binding free energy with respect to the active site resolution, the separate contributions coming from different energetic terms (such as electrostatic and van der Waals interactions) manifest a stronger dependence on the mapping, thus pointing to the existence of an optimal level of intermediate resolution.

A helical assembly of human ESCRT-I scaffolds reverse-topology membrane scission

The ESCRT complexes drive membrane scission in HIV-1 release, autophagosome closure, MVB biogenesis, cytokinesis, and other cell processes. ESCRT-I is the most upstream complex and bridges the system to HIV-1 Gag in virus release. The crystal structure of the headpiece of human ESCRT-I comprising TSG101–VPS28–VPS37B–MVB12A was determined, revealing an ESCRT-I helical assembly with a 12 molecule repeat. Electron microscopy confirmed that ESCRT-I subcomplexes form helical filaments in solution. Mutation of VPS28 helical interface residues blocks filament formation in vitro and autophagosome closure and HIV-1 release in human cells. Coarse grained simulations of ESCRT assembly at HIV-1 budding sites suggest that formation of a 12-membered ring of ESCRT-I molecules is a geometry-dependent checkpoint during late stages of Gag assembly and HIV-1 budding, and templates ESCRT-III assembly for membrane scission. These data show that ESCRT-I is not merely a bridging adaptor, but has an essential scaffolding and mechanical role in its own right.

Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.

Determining Optimal Coarse-Grained Representation for Biomolecules Using Internal Cluster Validation Indexes

The development of ultracoarse-grained models for large biomolecules needs to derive the optimal number of coarse-grained (CG) sites to represent the targets. In this work, we propose to use the statistical internal cluster validation indexes to determine the optimal number of CG sites that are optimized based on the essential dynamics coarse-graining method. The calculated curves of Calinski-Harabasz and Silhouette Coefficient indexes exhibit the extrema corresponding to the similar CG numbers. The calculated ratios of the optimal CG numbers to the residue numbers of fine-grained models are in the range from 4 to 2. The comparison of the stability of index results indicates that Calinski-Harabasz index is the better choice to determine the optimal CG representation in coarse-graining. © 2019 Wiley Periodicals, Inc.

A Multi-Scale Approach to Membrane Remodeling Processes

We present a multi-scale simulation procedure to describe membrane-related biological processes that span over a wide range of length scales. At macroscopic length-scale, a membrane is described as a flexible thin film modeled by a dynamic triangulated surface with its spatial conformations governed by an elastic energy containing only a few model parameters. An implicit protein model allows us to include complex effects of membrane-protein interactions in the macroscopic description. The gist of this multi-scale approach is a scheme to calibrate the implicit protein model using finer scale simulation techniques e.g., all atom and coarse grain molecular dynamics. We previously used this approach and properly described the formation of membrane tubular invaginations upon binding of B-subunit of Shiga toxin. Here, we provide a perspective of our multi-scale approach, summarizing its main features and sketching possible routes for future development.

Unusual Organization of I-BAR Proteins on Tubular and Vesicular Membranes

Protein-mediated membrane remodeling is a ubiquitous and critical process for proper cellular function. Inverse Bin/Amphiphysin/Rvs (I-BAR) domains drive local membrane deformation as a precursor to large-scale membrane remodeling. We employ a multiscale approach to provide the molecular mechanism of unusual I-BAR domain-driven membrane remodeling at a low protein surface concentration with near-atomistic detail. We generate a bottom-up coarse-grained model that demonstrates similar membrane-bound I-BAR domain aggregation behavior as our recent Mesoscopic Membrane with Explicit Proteins model. Together, these models bridge several length scales and reveal an aggregation behavior of I-BAR domains. We find that at low surface coverage (i.e., low bound protein density), I-BAR domains form transient, tip-to-tip strings on periodic flat membrane sheets. Inside of lipid bilayer tubules, we find linear aggregates parallel to the axis of the tubule. Finally, we find that I-BAR domains form tip-to-tip aggregates around the edges of membrane domes. These results are supported by in vitro experiments showing low curvature bulges surrounded by I-BAR domains on giant unilamellar vesicles. Overall, our models reveal new I-BAR domain aggregation behavior in membrane tubules and on the surface of vesicles at low surface concentration that add insight into how I-BAR domain proteins may contribute to certain aspects of membrane remodeling in cells.

A new method for the construction of coarse-grained models of large biomolecules from low-resolution cryo-electron microscopy data

The rapid development of cryo-electron microscopy (cryo-EM) has led to the generation of significant low-resolution electron density data of biomolecules. However, the atomistic details of huge biomolecules usually cannot be obtained because it is very difficult to construct all-atom models for MD simulations. Thus, it is still a challenge to make use of the rich low-resolution cryo-EM data for computer simulation and functional study. In this study, we proposed a new method called Convolutional and K-means Coarse-Graining (CK-CG) for the efficient coarse-graining of large biological systems. Using the CK-CG method, we could directly map the cryo-EM data into coarse-grained (CG) beads. Furthermore, the CG beads were parameterized with an empirical harmonic potential to construct a new CG model. We subjected the CK-CG models of the fibrillar protein assemblies F-actin and collagen to external forces in pulling dynamic simulations to assess their mechanical response. The agreement between the estimated tensile stiffness between CG models and experiments demonstrates the validity of the CK-CG method. Thus, our method provides a practical strategy for the direct construction of a structural model from low-resolution data for biological function studies.

Computational Modeling of Realistic Cell Membranes

Cell membranes contain a large variety of lipid types and are crowded with proteins, endowing them with the plasticity needed to fulfill their key roles in cell functioning. The compositional complexity of cellular membranes gives rise to a heterogeneous lateral organization, which is still poorly understood. Computational models, in particular molecular dynamics simulations and related techniques, have provided important insight into the organizational principles of cell membranes over the past decades. Now, we are witnessing a transition from simulations of simpler membrane models to multicomponent systems, culminating in realistic models of an increasing variety of cell types and organelles. Here, we review the state of the art in the field of realistic membrane simulations and discuss the current limitations and challenges ahead.

A deep learning approach to the structural analysis of proteins

Deep learning (DL) algorithms hold great promise for applications in the field of computational biophysics. In fact, the vast amount of available molecular structures, as well as their notable complexity, constitutes an ideal context in which DL-based approaches can be profitably employed. To express the full potential of these techniques, though, it is a prerequisite to express the information contained in a molecule’s atomic positions and distances in a set of input quantities that the network can process. Many of the molecular descriptors devised so far are effective and manageable for relatively small structures, but become complex and cumbersome for larger ones. Furthermore, most of them are defined locally, a feature that could represent a limit for those applications where global properties are of interest. Here, we build a DL architecture capable of predicting non-trivial and intrinsically global quantities, that is, the eigenvalues of a protein’s lowest-energy fluctuation modes. This application represents a first, relatively simple test bed for the development of a neural network approach to the quantitative analysis of protein structures, and demonstrates unexpected use in the identification of mechanically relevant regions of the molecule.

Backmapping from Multiresolution Coarse-Grained Models to Atomic Structures of Large Biomolecules by Restrained Molecular Dynamics Simulations Using Bayesian Inference

Coarse-grained (CG) simulations have allowed access to larger length scales and longer time scales in the study of the dynamic processes of large biomolecules than all-atom (AA) molecular dynamics (MD) simulations. Backmapping from CG models to AA structures has long been studied because it enables us to gain detailed structure insights from CG simulations. Many methods first construct an AA structure from the CG model by fragments, random placement, or geometrical rules and subsequently optimize the solution via energy minimization, simulated annealing or position-restrained simulations. However, such methods may only work well on residue-level CG models and cannot consider the deviations of CG models. In this work, we describe, to the best of our knowledge, a new backmapping method based on Bayesian inference and restrained MD simulations. Restraints with log harmonic energy terms are defined according to the target CG model using the Bayesian inference in which the CG deviations can be estimated. From an initial AA structure obtained from either high-resolution experiments or homology modeling, a MD simulation with the aforementioned restraints is performed to obtain a final AA structure that is a backmapping of the target CG model. The method was validated using multiresolution CG models of the soluble extracellular region of the human epidermal growth factor receptor and was further applied to construct AA structures from CG simulations of the nucleosome core particle. The results demonstrate that our method can generate accurate AA structures of different types of biomolecules from multiple CG models with either residue-level resolution or much lower resolution than one-site-per-residue.

Coarse-Grained Simulation of Full-Length Integrin Activation

Integrin conformational dynamics are critical to their receptor and signaling functions in many cellular processes, including spreading, adhesion, and migration. However, assessing integrin conformations is both experimentally and computationally challenging because of limitations in resolution and dynamic sampling. Thus, structural changes that underlie transitions between conformations are largely unknown. Here, focusing on integrin αvβ₃, we developed a modified form of the coarse-grained heterogeneous elastic network model (hENM), which allows sampling conformations at the onset of activation by formally separating local fluctuations from global motions. Both local fluctuations and global motions are extracted from all-atom molecular dynamics simulations of the full-length αvβ₃ bent integrin conformer, but whereas the former are incorporated in the hENM as effective harmonic interactions between groups of residues, the latter emerge by systematically identifying and treating weak interactions between long-distance domains with flexible and anharmonic connections. The new hENM model allows integrins and single-point mutant integrins to explore various conformational states, including the initiation of separation between α- and β-subunit cytoplasmic regions, headpiece extension, and legs opening.

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.