Pith. sign in

REVIEW 4 major objections 6 minor 81 references

Refining Coarse-Grained Molecular Topologies: A Bayesian Optimization Approach

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that refining only the bonded parameters of Martini3 topologies by Bayesian optimization—twelve to fifteen numbers per polymer—reproduces all-atom density and radius of gyration within about 10 percent error for all…

desk verdict Clever start/middle/end parametrization for Martini3 bonded terms, but the 'any degree of polymerization' claim is asserted, not tested—all reported errors are in-sample fits. read the letter →

arxiv 2501.02707 v4 pith:4R2MXT6L submitted 2025-01-06 physics.chem-ph cond-mat.mtrl-scics.LG

Pranoy Ray , Adam P. Generale , Nikhith Vankireddy , Yuichiro Asoma , Masataka Nakauchi , Haein Lee , Katsuhisa Yoshida , Yoshishige Okuno
show 1 more author
Surya R. Kalidindi
This is my paper · ORCID
classification physics.chem-phcond-mat.mtrl-scics.LG
keywords coarse-grainedmoleculardynamicsMartini3forcefieldBayesianoptimizationmulti-objectivepolymertopologydensityradiusofgyrationrefinement
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Coarse-grained simulations trade atomistic detail for speed, and the general-purpose Martini3 force field often misses the molecular-scale structure of a specific polymer. This paper argues that a small, targeted retuning of Martini3's bonded parameters—bond lengths, bond constants, angle magnitudes, angle constants, plus one aromatic bond length—can recover all-atom accuracy. Using Bayesian optimization with a multi-output Gaussian process and the q-EHVI acquisition function, the authors tune only the start, middle, and end segments of a polymer chain, cutting the parameter space from 139 to 15 dimensions for a 20-unit polystyrene chain. They report that for all 12 polymer systems (PE, PMMA, and PS at degrees of polymerization 3, 10, 20, and 50), the optimized coarse-grained simulations match all-atom density and radius of gyration within roughly 10 percent error, and they claim the resulting CG potential transfers to any degree of polymerization. If true, this gives a cheap way to build accurate coarse-grained models for domain-specific polymer applications.

What carries the argument

The load-bearing mechanism is a low-dimensional parametrization of the CG polymer topology: only the bonded parameters of the first, middle, and last segments (bond lengths $b_0$, bond constants $k_b$, angle magnitudes $\Phi$, angle constants $k_\Phi$, and, for aromatic polystyrene, an aromatic bond length $c$) are optimized, giving 12 or 15 design variables. The optimizer is a multi-output Gaussian process with an automatic-relevance-determination squared-exponential kernel and a linear model of coregionalization that jointly predicts density and radius of gyration, driven by the q-EHVI acquisition function, which balances exploration and exploitation while tracking the Pareto front between the two objectives. This machinery maps expensive CGMD runs to a tractable black-box optimization problem and is what lets the paper claim accuracy comparable to all-atom MD at CGMD cost.

What would settle it

Take the optimized start/middle/end parameters for, say, 10-PE and use them unchanged in a CGMD simulation of 50-PE; compare the computed density and radius of gyration to the corresponding all-atom values. If the absolute percentage errors exceed roughly 10 percent, the claim that the potential applies to any degree of polymerization is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a Martini3 topology refined by multi-objective Bayesian optimization can serve as a coarse-grained potential that is both accurate and transferable across chain lengths. The authors define molecular topology as the bonded parameters within a fixed CG mapping, and they compress the full bonded parameter set of a polymer chain into three groups—start, middle, and end segments—so that optimizing a 20-unit polystyrene chain requires only 15 numbers instead of 139. All-atom NPT simulations provide the reference values for density and radius of gyration; each CG parameter set is evaluated by running Martini3 simulations, and a Gaussian-process surrogate built with a linear model of coregionalization predicts both objectives so that q-EHVI can select the next batch of candidates. Iterating roughly 50 rounds (100 CGMD runs) drives the absolute percentage errors below about 10 percent for all 12 systems, with five independent seeds per polymer. The paper therefore proposes that the optimized start/middle/end parameters constitute a CG potential applicable to any degree of polymerization, a claim the authors put forward as a benchmark for future CG model building.

Load-bearing premise

The paper's transferability claim rests on the untested premise that parameters optimized for start, middle, and end segments at one degree of polymerization work at other chain lengths without retraining, since all validation is done at the same chain length used for optimization.

Editorial extensions

If this is right

  • If the optimized potentials transfer across chain lengths, one short optimization run per polymer class replaces per-length reparameterization.
  • The same loop can be pointed at experimental observables instead of all-atom reference values, since the objective function is built from abstract target properties.
  • The protocol's roughly 50-iteration convergence (about 100 CGMD runs, 3–11 hours on a single GPU) makes domain-specific Martini3 refinement affordable for routine materials screening.
  • Adding extra parameters to the start/middle/end scheme extends the approach to copolymers, branched chains, or polymers with side-group chemistry beyond PE, PMMA, and PS.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: The transferability claim is not yet tested because every reported result optimizes and evaluates at the same chain length; the assertion that the start/middle/end parameters work for any degree of polymerization is an extrapolation from that data.
  • Editorial inference: Because PMMA's density error grows with chain length and the paper attributes this to Martini3's non-bonded polar/nonpolar bead description, optimizing bonded terms alone may hit a ceiling for polar polymers; a combined bonded-plus-nonbonded refinement would be a testable extension.
  • Editorial inference: The start/middle/end compression implicitly assumes that polymer ends differ from the interior in a way that is independent of chain length; that assumption could be checked by computing full per-monomer parameters at several degrees of polymerization and comparing them to the optimized three-segment values.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The manuscript proposes a Bayesian-optimization workflow to refine Martini3 bonded parameters for coarse-grained polymer models. The key methodological idea is a low-dimensional parameterization in which bond length, bond constant, angle magnitude, and angle constant are optimized separately for the start, middle, and end segments of a polymer chain (plus an aromatic bond length for PS). The objective, Eq. (2), is the squared difference between CG and all-atom (AA) density and radius of gyration. For PE, PMMA, and PS at degrees of polymerization n=3, 10, 20, and 50, the authors report absolute percentage errors below about 10% relative to AAMD, show convergence curves and Pareto fronts, and claim that the resulting potential is applicable to any degree of polymerization. However, all reported results are in-sample: each of the 12 systems is optimized and evaluated at the same chain length, and no optimized parameter set from one n is transferred to another n. The central transferability claim is therefore not supported by the present experiments.

Significance. If the transferability claim were established, the workflow would be a practical and comparatively inexpensive way to specialize general-purpose Martini3 topologies for polymer families; the reported convergence in roughly 100 CGMD evaluations per system is plausible and is a useful practical data point. The paper also has clear strengths: the objective is explicitly defined, the optimization pipeline is described in enough detail to be reproduced, five seeds are used for error bars, and the comparison against raw Martini3 is systematic across 12 systems. Those features make the manuscript a potentially useful methodology demonstration. Nevertheless, the paper's principal claimed contribution, a validated 'CG potential applicable to any degree of polymerization,' is not supported by the current experimental design. The reported errors are training errors, and the proposed transfer mechanism is asserted rather than demonstrated. The significance of the work as submitted is therefore limited to per-system fitting, which is a considerably weaker claim than the one made in the Abstract and Results.

major comments (4)
  1. [Eq. (2); Results, Pareto Optimal Property Discrepancy Frontier; Figure 5] The reported errors below ~10% are in-sample training errors, not predictive accuracies. Eq. (2) defines the objective as the squared difference between CG and AA density and radius of gyration, and the workflow minimizes this objective separately for each of the 12 polymer systems. Figure 5 then reports the value of that same objective after optimization. There is no held-out chain length, no held-out parameter set, and no independent validation ensemble. Consequently, the statement that the framework 'reports exceptionally low errors under ~10% for both density and radius of gyration for all the 12 polymers studied' does not support a validated transferable potential; it supports only per-system fitting.
  2. [Low-dimensional Parametrization; Results; Abstract] The central claim of transferability across degrees of polymerization is not tested. The text states that the start/middle/end parameters 'are intended to be applicable when constructing models for other degrees of polymerization,' and the Abstract and Results assert a validated potential for 'any degree of polymerization.' However, every reported result is for chains where the parameters were optimized at that same n; no parameter set optimized at n=3 is ever simulated at n=50, and no interpolation or extrapolation in n is performed. A required addition is an explicit transfer test: optimize at one n (or on a training subset of n values) and report density and Rg errors at the remaining n values, with the raw Martini3 baseline included for comparison. The paper's own admission that a comparison to a uniform parameter set was not performed is also relevant: such a control is needed to establish that the three-segment parameterization, rather than simply the act of fitting, is responsible for any transferable improvement.
  3. [Table 1; Results; no table of optimized parameters] The optimized parameter values themselves are never reported. Since the paper claims to have developed a CG potential, a table of the final optimized values for all 12 systems (bond lengths, bond constants, angle magnitudes, angle constants, and the aromatic bond length c for PS) is necessary for independent verification, for use by other researchers, and for any meaningful claim of a reusable potential. Without these values, the reported error metrics cannot be checked, the parameters cannot be transferred by readers, and the 'potential' is not actually specified.
  4. [Results, Pareto Optimal Property Discrepancy Frontier; 50-PS discussion] The 50-PS result, where the raw Martini3 model matches experimental density better than the optimized model, exposes a limitation of using AAMD as the sole ground truth. Because the optimization target is the AA density, the workflow can minimize error relative to an AA reference that is itself in tension with experimental data. The authors acknowledge this point, but the central claim of 'accuracy comparable to AAMD' is then less informative unless experimental comparisons are also reported and the choice between AA-calibrated and experimentally calibrated targets is discussed. At minimum, the paper should state explicitly that the optimized potential is a fit to the AA model, not necessarily to experiment.
minor comments (6)
  1. [Low-dimensional Parametrization] The sentence 'Furthermore, this low-dimensional parametrization approach balances computational efficiency with the need to capture key topological features of the CG polymer' appears twice, at the end of the first and second paragraphs of the section; one copy should be removed.
  2. [Eq. (5)] The log marginal likelihood is missing a leading minus sign in the first term; as written, maximizing the expression would not correspond to the standard Gaussian process marginal likelihood, and the sign error would affect hyperparameter optimization.
  3. [Table 1] The PS start/end bond constant upper bound is listed as 90000, which is likely a typographical error (perhaps 9000), and the row for PE/PMMA angle constants uses k_b where k_phi is clearly intended.
  4. [Eqs. (11)-(12)] The notation for z_k^(m) and l_k^(m) is introduced too briefly; the superscripts, subscripts, and the min over candidate points should be defined precisely, and the text should state explicitly that Eq. (12) is an expanded Monte Carlo estimator of the same quantity defined in Eq. (11).
  5. [Figure 5] The error bars represent standard deviation over five seeds, but the text does not specify what varies between seeds (initial Latin Hypercube design, BO stochasticity, or MD initial velocities); this should be clarified so that the uncertainty is interpretable.
  6. [Methods and reproducibility] The manuscript does not state whether code, input files, example topologies, or final optimized parameter sets will be made available; a data and code availability statement is needed for reproducibility.

Circularity Check

2 steps flagged · score 7.0 of 10

Reported <10% errors are the optimized objective values from Eq. (2), and the 'any degree of polymerization' transferability is asserted without a cross-chain-length test.

  1. fitted input called prediction [Eq. (2), 'CG Topology Optimization Framework'; 'Pareto Optimal Property Discrepancy Frontier'; Fig. 5]
    "𝜽෡ = argmin 𝜃 ‖𝑘𝐶𝐺(𝜽) − 𝑘𝐴𝐴‖^2 (2) ... The models are trained until the objective function plateaus ... every polymer’s Martini3 topology is optimized until the absolute percentage errors converge to less than ~10%."

    The paper's headline accuracy numbers are the objective function in Eq. (2) evaluated at the optimized parameter set for the same degree of polymerization used during fitting. Each of the 12 polymer systems was optimized and evaluated at the same n, so Figure 5's density and radius-of-gyration errors are training losses, not out-of-sample predictions. The reported '<10%' accuracy is therefore, by the paper's own equation, the convergence target of the optimizer rather than an independent validation of the resulting CG topology.

  2. other [Abstract; Results, 'Low-dimensional Parametrization of the Coarse-grained Molecular Structure']
    "We have developed and validated a CG potential applicable to any degree of polymerization ... The parameters optimized for these regions are intended to be applicable when constructing models for other degrees of polymerization of the same polymer, thus promoting transferability."

    The 'any degree of polymerization' claim is not derived from any cross-n experiment: parameters are optimized separately at n = 3, 10, 20, and 50 and evaluated at the same n. No parameter set optimized at one n is simulated at another n. The start/middle/end ansatz is asserted as the mechanism for transferability, but the paper's own language—'intended to be applicable'—concedes that transferability is assumed rather than tested. The only evidence offered for the universality claim is the in-sample fits, so the validation statement reduces to the training data.

full rationale

The derivation chain begins with Eq. (2), which defines the optimized bonded parameters as the minimizer of the squared difference between CG and AA density and radius of gyration. The paper then reports that optimization converged to '<10%' errors for all 12 systems and presents Figure 5 as evidence of accuracy. By construction, these are the values of the minimized objective, not independent predictions: the parameters were fit to those same properties at those same chain lengths. This is a clear instance of a fitted quantity being reported as a validated prediction. The broader claim that the optimized potential is 'applicable to any degree of polymerization' is load-bearing but untested; optimization was performed per n, and no hold-out n was ever simulated. The paper's admission that no uniform-parameter baseline was run reinforces the absence of a transferability check. The BO framework itself is not circular—it uses external AA references and is a legitimate fitting procedure—so the score reflects partial circularity of the central accuracy claim plus an unsupported universality claim, rather than a fully self-citation-driven derivation.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central claim rests on several fitted bonded parameters and on assumptions about the sufficiency of the low-dimensional parametrization and the correctness of the AAMD reference. No new physical entities are introduced. The paper does not report the fitted parameter values, which limits independent assessment.

free parameters (5)
  • Bond length b0 for start, middle, and end segments = not reported
    Optimized via BO per polymer and per segment to match AAMD density and Rg.
  • Bond constant kb for start, middle, and end segments = not reported
    Optimized via BO per polymer and per segment to match AAMD density and Rg.
  • Angle magnitude Phi for start, middle, and end segments = not reported
    Optimized via BO per polymer and per segment to match AAMD density and Rg.
  • Angle constant kPhi for start, middle, and end segments = not reported
    Optimized via BO per polymer and per segment to match AAMD density and Rg.
  • Aromatic bond length c for PS = not reported
    Additional parameter for polystyrene, optimized via BO to match AAMD density and Rg.
assumptions (3)
  • domain assumption The bonded parameters of the start, middle, and end segments sufficiently determine density and radius of gyration for all degrees of polymerization.
    Introduced in the Low-dimensional Parametrization section; the paper explicitly notes no comparison to a uniform parameter set was performed.
  • domain assumption AAMD (GAFF) simulations provide the correct ground-truth density and radius of gyration.
    Used as reference in Eq. (2); the paper later notes Martini3 matches experimental density better than the optimized model for 50-PS, undermining this ground-truth assumption.
  • domain assumption Optimizing only bonded parameters, while holding non-bonded Martini3 parameters fixed, can bring CGMD density and Rg to within 10 percent of AAMD.
    The optimization space in Table 1 includes only bonded terms, while density is acknowledged to depend on non-bonded interactions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Refining Coarse-Grained Molecular Topologies: A Bayesian Optimization Approach." pith.science (2026). https://pith.science/paper/4R2MXT6L

@misc{pith2026250102707,
  author       = {Pith},
  title        = {Pith review of: Refining Coarse-Grained Molecular Topologies: A Bayesian Optimization Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4R2MXT6L}},
  note         = {Machine review of arXiv:2501.02707}
}
read the original abstract

Molecular Dynamics (MD) simulations are essential for accurately predicting the physical and chemical properties of large molecular systems across various pressure and temperature ensembles. However, the high computational costs associated with All-Atom (AA) MD simulations have led to the development of Coarse-Grained Molecular Dynamics (CGMD), providing a lower-dimensional compression of the AA structure into representative CG beads, offering reduced computational expense at the cost of predictive accuracy. Existing CGMD methods, such as CG-Martini (calibrated against experimental data), aim to generate an embedding of a topology that sufficiently generalizes across a range of structures. Detrimentally, in attempting to specify parameterization with applicability across molecular classes, it is unable to specialize to domain-specific applications, where sufficient accuracy and computational speed are critical. This work presents a novel approach to optimize derived results from CGMD simulations by refining the general-purpose Martini3 topologies specifically the bonded interaction parameters within a given coarse-grained mapping - for domain-specific applications using Bayesian Optimization methodologies. We have developed and validated a CG potential applicable to any degree of polymerization, representing a significant advancement in the field. Our optimized CG potential, based on the Martini3 framework, aims to achieve accuracy comparable to AAMD while maintaining the computational efficiency of CGMD. This approach bridges the gap between efficiency and accuracy in multiscale molecular simulations, potentially enabling more rapid and cost-effective molecular discovery across various scientific and technological domains.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

81 extracted references · 71 canonical work pages

  1. [1]

    & Hall, C

    Voegler Smith, A. & Hall, C. K. α‐Helix formation: Discontinuous molecular dynamics on an intermediate‐resolution protein model. Proteins Struct. Funct. Bioinforma. 44, 344–360 (2001)

    2001

  2. [2]

    M., Buldyrey, S

    Ding, F., Borreguero, J. M., Buldyrey, S. V., Stanley, H. E. & Dokholyan, N. V. Mechanism for the α‐helix to β‐hairpin transition. Proteins Struct. Funct. Bioinforma. 53, 220–228 (2003)

    work page 2003

  3. [3]

    Gartner, T. E. & Jayaraman, A. Modeling and Simulations of Polymers: A Roadmap. Macromolecules 52, 755–786 (2019)

    work page 2019

  4. [4]

    M., Huang, W

    Ravikumar, K. M., Huang, W. & Yang, S. Coarse-Grained Simulations of Protein-Protein Association: An Energy Landscape Perspective. Biophys. J. 103, 837–845 (2012)

    work page 2012

  5. [5]

    P., Acuto, O

    Prakaash, D., Fagnen, C., Cook, G. P., Acuto, O. & Kalli, A. C. Molecular dynamics simulations reveal membrane lipid interactions of the full-length lymphocyte specific kinase (Lck). Sci. Rep. 12, 21121 (2022)

    work page 2022

  6. [6]

    Alder, B. J. & Wainwright, T. E. Studies in Molecular Dynamics. I. General Method. J. Chem. Phys. 31, 459–466 (1959)

    work page 1959

  7. [7]

    B., Goland, A

    Gibson, J. B., Goland, A. N., Milgram, M. & Vineyard, G. H. Dynamics of Radiation Damage. Phys. Rev. 120, 1229–1253 (1960). 13

    work page 1960

  8. [8]

    Correlations in the Motion of Atoms in Liquid Argon

    Rahman, A. Correlations in the Motion of Atoms in Liquid Argon. Phys. Rev. 136, A405– A411 (1964)

    work page 1964

Show all 81 references
  1. [9]

    & Banerjee, S

    Chakraborty, B., Ray, P., Garg, N. & Banerjee, S. High capacity reversible hydrogen storage in titanium doped 2D carbon allotrope Ψ-graphene: Density Functional Theory investigations. Int. J. Hydrog. Energy 46, 4154–4167 (2021)

    2021

  2. [10]

    & Chakraborty, B

    Kundu, A., Jaiswal, A., Ray, P., Sahu, S. & Chakraborty, B. Zr doped C 24 fullerene as efficient hydrogen storage material: insights from DFT simulations. J. Phys. Appl. Phys. 57, 495502 (2024)

    2024

  3. [11]

    T., Kundu, A., Ray, P., Jha, P

    Nair, H. T., Kundu, A., Ray, P., Jha, P. K. & Chakraborty, B. Ti-decorated C30 as a High- capacity Hydrogen Storage Material: Insights from Density Functional Theory. Sustain. Energy Fuels (2023) doi:10.1039/D3SE00845B

    2023 doi

  4. [12]

    S., Kowalik, M., Jensen, B

    Gaikwad, P. S., Kowalik, M., Jensen, B. D., Van Duin, A. & Odegard, G. M. Molecular Dynamics Modeling of Interfacial Interactions between Flattened Carbon Nanotubes and Amorphous Carbon: Implications for Ultra-Lightweight Composites. ACS Appl. Nano Mater. 5, 5915–5924 (2022)

    2022

  5. [13]

    J., Risselada, H

    Marrink, S. J., Risselada, H. J., Yefimov, S., Tieleman, D. P. & De Vries, A. H. The MARTINI Force Field: Coarse Grained Model for Biomolecular Simulations. J. Phys. Chem. B 111, 7812–7824 (2007)

    2007

  6. [14]

    & Müller‐Plathe, F

    Reith, D., Pütz, M. & Müller‐Plathe, F. Deriving effective mesoscale potentials from atomistic simulations. J. Comput. Chem. 24, 1624–1636 (2003)

    2003

  7. [15]

    Coarse-Graining in Polymer Simulation: From the Atomistic to the Mesoscopic Scale and Back

    Müller-Plathe, F. Coarse-Graining in Polymer Simulation: From the Atomistic to the Mesoscopic Scale and Back. Chemphyschem Eur. J. Chem. Phys. Phys. Chem. 3, 755–69 (2002). 14

    2002

  8. [16]

    & Voth, G

    Izvekov, S. & Voth, G. A. A Multiscale Coarse-Graining Method for Biomolecular Systems. J. Phys. Chem. B 109, 2469–2473 (2005)

    2005

  9. [17]

    Noid, W. G. et al. The multiscale coarse-graining method. I. A rigorous bridge between atomistic and coarse-grained models. J. Chem. Phys. 128, 244114 (2008)

    2008

  10. [18]

    Shell, M. S. The relative entropy is fundamental to multiscale and inverse thermodynamic problems. J. Chem. Phys. 129, 144108 (2008)

    2008

  11. [19]

    Penfold, N. J. W., Yeow, J., Boyer, C. & Armes, S. P. Emerging Trends in Polymerization- Induced Self-Assembly. ACS Macro Lett. 8, 1029–1054 (2019)

    2019

  12. [20]

    Borges-Araújo, L. et al. Martini 3 Coarse-Grained Force Field for Cholesterol. J. Chem. Theory Comput. 19, 7387–7404 (2023)

    2023

  13. [21]

    Risselada, H. J. Martini 3: a coarse-grained force field with an eye for atomic detail. Nat. Methods 18, 342–343 (2021)

    2021

  14. [22]

    Souza, P. C. T. et al. Martini 3: a general purpose force field for coarse-grained molecular dynamics. Nat. Methods 18, 382–388 (2021)

    2021

  15. [23]

    Alessandri, R. et al. Martini 3 Coarse-Grained Force Field: Small Molecules. Preprint at https://doi.org/10.26434/chemrxiv-2021-1qmq9 (2021)

    2021 doi

  16. [24]

    Brosz, M., Michelarakis, N., Bunz, U. H. F., Aponte-Santamaría, C. & Gräter, F. Martini 3 coarse-grained force field for poly( para -phenylene ethynylene)s. Phys. Chem. Chem. Phys. 24, 9998–10010 (2022)

    2022

  17. [25]

    & Courville, A

    Goodfellow, I., Bengio, Y. & Courville, A. Deep Learning. (The MIT Press, Cambridge, Massachusetts, 2016)

    2016

  18. [26]

    & Kalidindi, S

    Ray, P., Choudhary, K. & Kalidindi, S. R. Lean CNNs for Mapping Electron Charge Density Fields to Material Properties. Integrating Mater. Manuf. Innov. 14, 1–13 (2025). 15

    2025

  19. [27]

    & Andrienko, D

    Rühle, V., Junghans, C., Lukyanov, A., Kremer, K. & Andrienko, D. Versatile Object- Oriented Toolkit for Coarse-Graining Applications. J. Chem. Theory Comput. 5, 3211–3223 (2009)

    2009

  20. [28]

    & Lyubartsev, A

    Mirzoev, A. & Lyubartsev, A. P. MagiC: Software Package for Multiscale Modeling. J. Chem. Theory Comput. 9, 1512–1520 (2013)

    2013

  21. [29]

    & Zavadlav, J

    Fuchs, P., Thaler, S., Röcken, S. & Zavadlav, J. chemtrain: Learning deep potential models via automatic differentiation and statistical physics. Comput. Phys. Commun. 310, 109512 (2025)

    2025

  22. [30]

    Wang, X. et al. DMFF: An Open-Source Automatic Differentiable Platform for Molecular Force Field Development and Molecular Dynamics Simulation. J. Chem. Theory Comput. 19, 5897–5909 (2023)

    2023

  23. [31]

    TORCHMD-NET: EQUIVARIANT TRANSFORMERS FOR NEURAL NETWORK BASED MOLECULAR POTENTIALS

    Tholke, P. TORCHMD-NET: EQUIVARIANT TRANSFORMERS FOR NEURAL NETWORK BASED MOLECULAR POTENTIALS. (2022)

    2022

  24. [32]

    Zeng, J. et al. DeePMD-kit v2: A software package for deep potential models. J. Chem. Phys. 159, 054801 (2023)

    2023

  25. [33]

    Peng, Y. et al. OpenMSCG: A Software Tool for Bottom-Up Coarse-Graining. J. Phys. Chem. B 127, 8537–8550 (2023)

    2023

  26. [34]

    Abraham, M. J. et al. GROMACS: High performance molecular simulations through multi- level parallelism from laptops to supercomputers. SoftwareX 1–2, 19–25 (2015)

    2015

  27. [35]

    Thompson, A. P. et al. LAMMPS - a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales. Comput. Phys. Commun. 271, 108171 (2022). 16

    2022

  28. [36]

    & Kremer, K

    Bereau, T. & Kremer, K. Automated Parametrization of the Coarse-Grained Martini Force Field for Small Organic Molecules. J. Chem. Theory Comput. 11, 2783–2791 (2015)

    2015

  29. [37]

    Pereira, G. P. et al. Bartender: Martini 3 Bonded Terms via Quantum Mechanics-Based Molecular Dynamics. J. Chem. Theory Comput. 20, 5763–5773 (2024)

    2024

  30. [38]

    Mishra, A. et al. Multiobjective genetic training and uncertainty quantification of reactive force fields. Npj Comput. Mater. 4, 42 (2018)

    2018

  31. [39]

    Chan, H. et al. Machine learning coarse grained models for water. Nat. Commun. 10, 379 (2019)

    2019

  32. [40]

    Empereur-Mot, C. et al. Swarm-CG : Automatic Parametrization of Bonded Terms in MARTINI-Based Coarse-Grained Models of Simple to Complex Molecules via Fuzzy Self- Tuning Particle Swarm Optimization. ACS Omega 5, 32823–32843 (2020)

    2020

  33. [41]

    S., Souza, P

    Stroh, K. S., Souza, P. C. T., Monticelli, L. & Risselada, H. J. CGCompiler: Automated Coarse-Grained Molecule Parametrization via Noise-Resistant Mixed-Variable Optimization. J. Chem. Theory Comput. 19, 8384–8400 (2023)

    2023

  34. [42]

    K., Engel, M

    Krueger, R. K., Engel, M. C., Hausen, R. & Brenner, M. P. Fitting Coarse-Grained Models to Macroscopic Experimental Data via Automatic Differentiation. Preprint at https://doi.org/10.48550/arXiv.2411.09216 (2025)

    2025 doi arXiv

  35. [43]

    & Zhou, T

    Wu, Z. & Zhou, T. Structural Coarse-Graining via Multiobjective Optimization with Differentiable Simulation. J. Chem. Theory Comput. 20, 2605–2617 (2024)

    2024

  36. [44]

    Bayesian Approach to Global Optimization: Theory and Applications

    Mockus, J. Bayesian Approach to Global Optimization: Theory and Applications. (Springer Netherlands, Dordrecht, 1989). 17

    1989

  37. [45]

    On bayesian methods for seeking the extremum

    Močkus, J. On bayesian methods for seeking the extremum. in Optimization Techniques IFIP Technical Conference Novosibirsk, July 1–7, 1974 (ed. Marchuk, G. I.) vol. 27 400–404 (Springer Berlin Heidelberg, Berlin, Heidelberg, 1975)

    1974

  38. [46]

    Wu, Y., Walsh, A. & M. Ganose, A. Race to the bottom: Bayesian optimisation for chemical problems. Digit. Discov. 3, 1086–1100 (2024)

    2024

  39. [47]

    van Henten, G. B. et al. Comparison of optimization algorithms for automated method development of gradient profiles. J. Chromatogr. A 1742, 465626 (2025)

    2025

  40. [48]

    Moeini, M., Sela, L., Taha, A. F. & Abokifa, A. A. Optimization Techniques for Chlorine Dosage Scheduling in Water Distribution Networks: A Comparative Analysis. in World Environmental and Water Resources Congress 2023 987–998 (American Society of Civil Engineers, Henderson, N...

    2023 doi

  41. [49]

    & Adachi, Y

    Wang, Z., Ogawa, T. & Adachi, Y. Influence of Algorithm Parameters of Bayesian Optimization, Genetic Algorithm, and Particle Swarm Optimization on Their Optimization Performance. Adv. Theory Simul. 2, 1900110 (2019)

    2019

  42. [50]

    & Bochicchio, D

    Cambiaso, S., Rasera, F., Rossi, G. & Bochicchio, D. Development of a transferable coarse- grained model of polydimethylsiloxane. Soft Matter 18, 7887–7896 (2022)

    2022

  43. [51]

    & Sadowski, G

    Fischer, J., Paschek, D., Geiger, A. & Sadowski, G. Modeling of Aqueous Poly(oxyethylene) Solutions. 2. Mesoscale Simulations. J. Phys. Chem. B 112, 13561–13571 (2008)

    2008

  44. [52]

    & Theodorou, D

    Kamio, K., Moorthi, K. & Theodorou, D. N. Coarse Grained End Bridging Monte Carlo Simulations of Poly(ethylene terephthalate) Melt. Macromolecules 40, 710–722 (2007)

    2007

  45. [53]

    Alvares, C. M. S., Maurin, G. & Semino, R. Coarse-grained modeling of zeolitic imidazolate framework-8 using MARTINI force fields. J. Chem. Phys. 158, 194107 (2023). 18

    2023

  46. [54]

    Alessandri, R. et al. Pitfalls of the Martini Model. J. Chem. Theory Comput. 15, 5448–5460 (2019)

    2019

  47. [55]

    R., Gul, E

    Joseph, V. R., Gul, E. & Ba, S. Maximum projection designs for computer experiments. Biometrika 102, 371–380 (2015)

    2015

  48. [56]

    & Bakshy, E

    Daulton, S., Balandat, M. & Bakshy, E. Parallel Bayesian Optimization of Multiple Noisy Objectives with Expected Hypervolume Improvement. Preprint at https://doi.org/10.48550/ARXIV.2105.08195 (2021)

    2021 doi arXiv

  49. [57]

    & Smiatek, J

    Vögele, M., Holm, C. & Smiatek, J. Coarse-grained simulations of polyelectrolyte complexes: MARTINI models for poly(styrene sulfonate) and poly(diallyldimethylammonium). J. Chem. Phys. 143, 243151 (2015)

    2015

  50. [58]

    & Wang, Z

    Lin, K. & Wang, Z. Multiscale mechanics and molecular dynamics simulations of the durability of fiber-reinforced polymer composites. Commun. Mater. 4, 66 (2023)

    2023

  51. [59]

    Hollingsworth, S. A. & Dror, R. O. Molecular Dynamics Simulation for All. Neuron 99, 1129–1143 (2018)

    2018

  52. [60]

    Gulrez, S. K. H. et al. A review on electrically conductive polypropylene and polyethylene. Polym. Compos. 35, 900–914 (2014)

    2014

  53. [61]

    & Pötschke, P

    Krause, B., Boldt, R., Häußler, L. & Pötschke, P. Ultralow percolation threshold in polyamide 6.6/MWCNT composites. Compos. Sci. Technol. 114, 119–125 (2015)

    2015

  54. [62]

    Liu, X., Li, C., Pan, Y., Schubert, D. W. & Liu, C. Shear-induced rheological and electrical properties of molten poly(methyl methacrylate)/carbon black nanocomposites. Compos. Part B Eng. 164, 37–44 (2019)

    2019

  55. [63]

    (Springer Singapore, Singapore, 2016)

    Computer Simulation of Polymeric Materials: Applications of the OCTA System. (Springer Singapore, Singapore, 2016). doi:10.1007/978-981-10-0815-3. 19

    2016 doi

  56. [64]

    Wang, J., Wang, W., Kollman, P. A. & Case, D. A. Automatic atom type and bond type perception in molecular mechanical calculations. J. Mol. Graph. Model. 25, 247–260 (2006)

    2006

  57. [65]

    M., Caldwell, J

    Wang, J., Wolf, R. M., Caldwell, J. W., Kollman, P. A. & Case, D. A. Development and testing of a general amber force field. J. Comput. Chem. 25, 1157–1174 (2004)

    2004

  58. [66]

    Frisch, M. J. et al. Gaussian˜16 Revision B.01. (2016)

    2016

  59. [67]

    Kroon, P. et al. Martinize2 and Vermouth: Unified Framework for Topology Generation. Preprint at https://doi.org/10.7554/eLife.90627.2 (2024)

    2024 doi

  60. [68]

    De Jong, D. H. et al. Improved Parameters for the Martini Coarse-Grained Protein Force Field. J. Chem. Theory Comput. 9, 687–697 (2013)

    2013

  61. [69]

    Talapatra, A. et al. Autonomous efficient experiment design for materials discovery with Bayesian model averaging. Phys. Rev. Mater. 2, 113803 (2018)

    2018

  62. [70]

    & Arroyave, R

    Honarmandi, P., Attari, V. & Arroyave, R. Accelerated materials design using batch Bayesian optimization: A case study for solving the inverse problem from materials microstructure to process specification. Comput. Mater. Sci. 210, 111417 (2022)

    2022

  63. [71]

    Hu, R. et al. Protein engineering via Bayesian optimization-guided evolutionary algorithm and robotic experiments. Brief. Bioinform. 24, bbac570 (2023)

    2023

  64. [72]

    Deneault, J. R. et al. Toward autonomous additive manufacturing: Bayesian optimization on a 3D printer. MRS Bull. 46, 566–575 (2021)

    2021

  65. [73]

    Du, L., Gao, R., Suganthan, P. N. & Wang, D. Z. W. Bayesian optimization based dynamic ensemble for time series forecasting. Inf. Sci. 591, 155–175 (2022)

    2022

  66. [74]

    MacKay, D. J. C. Information Theory, Inference, and Learning Algorithms. (Cambridge University Press, Cambridge, 2019). 20

    2019

  67. [75]

    Frazier, P. I. A Tutorial on Bayesian Optimization. Preprint at https://doi.org/10.48550/ARXIV.1807.02811 (2018)

    2018 doi arXiv

  68. [76]

    Rasmussen, C. E. & Williams, C. K. I. Gaussian Processes for Machine Learning. (The MIT Press, 2005). doi:10.7551/mitpress/3206.001.0001

    2005 doi

  69. [77]

    Bishop, C. M. Pattern Recognition and Machine Learning. (Springer, New York, 2006)

    2006

  70. [78]

    Machine Learning - A Probabilistic Perspective

    Murphy, K. Machine Learning - A Probabilistic Perspective. (MIT Press, Cambridge, 2014)

    2014

  71. [79]

    A., Rosasco, L

    Alvarez, M. A., Rosasco, L. & Lawrence, N. D. Kernels for Vector-Valued Functions: a Review. Preprint at https://doi.org/10.48550/arXiv.1106.6251 (2012)

    2012 doi arXiv

  72. [80]

    Journel, A. G. & Huijbregts, C. J. Mining Geostatistics. (Blackburn Press, Caldwell, N.J, 2003)

    2003

  73. [81]

    T., Moriconi, R., Hutter, F

    Wilson, J. T., Moriconi, R., Hutter, F. & Deisenroth, M. P. The reparameterization trick for acquisition functions. Preprint at https://doi.org/10.48550/arXiv.1712.00424 (2017). FIGURE LEGENDS Figure 1: Low-dimensional parametrization of a coarse-grained polymer chain. The sch...

    2017 doi arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.