Pith. sign in

REVIEW 3 major objections 4 minor 119 references

An implicit-solvent machine-learning potential trained only on ab initio and experimental labels matches explicit-solvent accuracy for drugs and proteins while evaluating timesteps two orders of magnitude faster.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-14 08:31 UTC pith:6Y5P3PB7

load-bearing objection Solid multiscale prior-free implicit-solvent MLP that actually transfers to proteins and beats prior ML CG models; the classical-config force-matching step is a real but partially mitigated soft spot, not a collapse of the claim. the 3 major comments →

arxiv 2607.10887 v1 pith:6Y5P3PB7 submitted 2026-07-12 physics.chem-ph cs.LGq-bio.BM

Transferable Implicit Solvent Machine Learning Potential for Drugs and Proteins Approaching Ab Initio Accuracy

Jan Eckwert , Julija Zavadlav This is my paper
classification physics.chem-ph cs.LGq-bio.BM
keywords machine learning potentialsimplicit solventequivariant graph neural networksforce matchingsolvation free energyprotein dynamicsdrug-like molecules
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Atomistic machine-learning potentials can approach quantum accuracy, but their cost still blocks the microsecond-scale sampling that biomolecules require. Implicit-solvent models remove the water molecules and therefore run far faster, yet earlier machine-learning versions were trained on classical force-field trajectories and therefore inherited those force fields' errors. This paper introduces TWIN, an equivariant graph neural network that learns the effective water-mediated interactions of a solute directly from density-functional data and experimental hydration free energies, without any classical force-field scaffold. Across drug-like molecules, peptides and folded proteins the model recovers free-energy landscapes, crystallographic structures and NMR observables that match both its own explicit-solvent counterpart and experiment, while each molecular-dynamics step is roughly one hundred times cheaper. The result is a practical route to ab-initio-level aqueous simulations of systems that were previously out of reach.

Core claim

TWIN, a fully data-driven equivariant graph neural network trained solely on ab initio forces and experimental solvation free energies, transfers across drug-like molecules, peptides and proteins, reproduces free-energy surfaces of its explicit-solvent DFT-based counterpart and experimental crystallographic and NMR benchmarks, and evaluates each molecular-dynamics timestep two orders of magnitude faster.

What carries the argument

A three-stage multiscale training pipeline: atomistic pre-training on DFT labels (TWIN-AT), variational force-matching of solvent-averaged forces on protein configurations (TWIN-FM), and top-down free-energy-path reweighting against experimental hydration free energies (final TWIN).

Load-bearing premise

That the bias introduced by force-matching labels on configurations generated by a classical force field, rather than by the atomistic machine-learning model itself, can be fully removed by the later free-energy refinement step.

What would settle it

A free-energy surface or NMR order-parameter set for a held-out peptide or protein that systematically disagrees with both the explicit-solvent TWIN-AT reference and experiment after the three-stage training is complete.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Microsecond-scale aqueous simulations of proteins and protein–ligand complexes become feasible at near-DFT accuracy without classical force-field priors.
  • The same multiscale recipe can be reapplied to other solvents once corresponding free-energy data exist.
  • Long-range electrostatic corrections and more diverse unfolded training configurations can be added without redesigning the architecture.
  • TWIN energies can serve as labels for coarse-grained generative models of biomolecular ensembles.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Because the model never relies on a topology-dependent prior, it can in principle be applied to non-natural amino acids and covalent modifications without re-parameterization.
  • The residual over-flexibility seen in order parameters suggests that adding explicit long-range electrostatics would simultaneously tighten structures and improve distance-dependent NMR observables.
  • If the intermediate classical-configuration bias proves harder to correct for larger or more disordered systems, the practical domain of the method will remain limited to near-native ensembles.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The manuscript introduces TWIN, a MACE-based implicit-solvent machine-learning potential for aqueous biomolecular systems. It is trained in three stages: (i) bottom-up energy/force matching of an explicit-solvent atomistic model (TWIN-AT) on a curated SPICE-derived DFT dataset, (ii) variational force matching of TWIN-AT forces on 4.1 million classical AMBER+TIP3P configurations of 41 CATH domains to obtain TWIN-FM, and (iii) top-down ReSolv refinement against experimental hydration free energies from CombiSolv/FreeSolv. The authors claim that the resulting model, free of classical force-field priors, transfers across drug-like molecules, peptides and proteins, reproduces free-energy landscapes of its explicit-solvent counterpart to near-ab-initio accuracy, matches crystallographic and NMR observables at a level competitive with or better than prior ML implicit/CG models and classical force fields, and evaluates timesteps one-to-two orders of magnitude faster.

Significance. If the central claims hold, TWIN would constitute a genuine advance: an equivariant-GNN implicit solvent model trained exclusively on ab-initio and experimental labels that simultaneously achieves transferability across chemical space and molecular size, quantitative free-energy agreement with a DFT-based explicit-solvent MLP, and the computational efficiency needed for microsecond-scale biomolecular sampling. The multiscale training strategy (pre-training + force matching + ReSolv) and the deliberate avoidance of topology-dependent priors are technically interesting and address a recognized data bottleneck in bottom-up CG/implicit MLPs. Extensive external benchmarks (CCSD biaryl torsions, NOE violations, J-couplings, hydrogen-bond scalar couplings, order parameters, COSMO-RS solvation shifts) strengthen the case. These strengths merit serious consideration provided residual biases from the formally inconsistent force-matching stage can be quantified.

major comments (3)
  1. [Results, second training stage; Methods, TWIN-FM] Results (second training stage) and Methods (TWIN-FM model): Force matching is performed on 4.1 M configurations generated by AMBER+TIP3P and labeled by TWIN-AT. The paper explicitly states that vanishing effective sample size precludes reweighting, producing a formal inconsistency whose bias is only later corrected by ReSolv on small-molecule hydration free energies. For Ala3 and ozanimod the third stage visibly improves the PMF toward TWIN-AT (Supplementary Figs. 5, 7), yet residual basin shifts remain and the FreeSolv/CombiSolv chemistry contains almost no protein-scale or highly flexible macrocycle content. The protein NMR/crystallographic agreement and the claim of “approaching ab initio accuracy” therefore rest on an unquantified residual classical bias. A concrete quantification (e.g., free-energy differences or force residuals on held-out protein fragments after ReSolv, or a rewe
  2. [Results, Proteins] Results (Proteins section): All four protein benchmarks (ubiquitin, GB1, CspA, IFABP) are limited to 10 ns unbiased trajectories. While RMSD, RMSF, h3JNC, S2 and lysine methyl order parameters look competitive with Amber, the paper itself notes that lysine rotamers interconvert slowly and that 3JCγNζ is highly sensitive to rotamer populations. Ten nanoseconds is insufficient to establish that the effective PMF has removed classical bias or that long-timescale stability matches the explicit-solvent reference. Longer trajectories or enhanced-sampling free-energy comparisons on at least one protein-sized system are needed to support the transferability and accuracy claims at the protein scale.
  3. [Abstract; Discussion] Abstract / Discussion / title claim of “approaching Ab Initio Accuracy”: For drug-like molecules and Ala3 the paper supplies direct free-energy comparisons to TWIN-AT (and, for ozanimod, to ANI-1ccx). For proteins the comparison is only to experiment and to classical force fields; direct free-energy or force residuals versus TWIN-AT are computationally infeasible and therefore absent. Given the acknowledged formal inconsistency of stage 2, the unqualified “ab initio” language over-reaches for protein systems. The claim should be restricted to the systems for which explicit free-energy agreement with TWIN-AT is demonstrated, or additional diagnostics supplied.
minor comments (4)
  1. [Results, Fig. 1c] Figure 1 caption and main text: the FreeSolv MAE of 0.96 kcal mol−1 is reported after noting that FreeSolv is “partly represented” in CombiSolv; a clear statement of the overlap size (or a fully held-out FreeSolv subset) would strengthen the solvation-thermodynamics claim.
  2. [Methods, Eq. (1)] Methods (TWIN-AT): the change of loss weights from (ωF=1e3, ωU=40) to (ωF=10, ωU=1e3) is stated without justification or ablation; a short sentence on why the schedule was chosen would aid reproducibility.
  3. [Results, Computational Performance] Computational Performance: speed-ups are given relative to TWIN-AT on A100 hardware with chemtrain-deploy; a brief comparison against a production classical implicit-solvent implementation (e.g., Amber OBC2 in OpenMM) under identical settings would place the absolute performance in clearer context.
  4. [Results, Proteins] Several Supplementary Figure references (e.g., Supplementary Fig. 7 for protein RMSD) are cited in the main text before the corresponding experimental observables are fully defined; reordering or a short clarifying sentence would improve readability.

Circularity Check

1 steps flagged

Mild fitted-input reporting on CombiSolv free energies; core transferability and ab-initio-matching claims rest on independent NMR/X-ray and TWIN-AT free-energy benchmarks.

specific steps
  1. fitted input called prediction [Results, final training stage / Fig. 1c]
    "TWIN predicts hydration free energies for the aqueous CombiSolv subset with a mean absolute error (MAE) of 0.76 kcal/mol (Figure 1c). Evaluation on FreeSolv, a commonly used hydration-free-energy benchmark that is partly represented in the experimental CombiSolv curation, yields a MAE of 0.96 kcal/mol. ... These results demonstrate state-of-the-art performance in solvation thermodynamics across MLPs and classical force fields."

    CombiSolv is the exact top-down fine-tuning set (Methods: “fine-tuning it on experimental hydration free energies from the CombiSolv dataset”; loss L(θ)=(ΔA(θ)−ΔA^exp)^2). The reported MAE is therefore the training residual (and FreeSolv is acknowledged as partly overlapping), not an independent out-of-sample prediction. The quantity being “predicted” is statistically forced by the ReSolv objective.

full rationale

The multiscale procedure (TWIN-AT force/energy matching on SPICE DFT, TWIN-FM variational force matching of TWIN-AT labels on classical CATH configurations, ReSolv top-down on CombiSolv experimental hydration free energies) is a standard supervised + reweighting pipeline, not a closed mathematical derivation. Validation targets (ozanimod/Ala3 PMFs vs TWIN-AT, protein RMSD/RMSF/h3JNC/S2 vs crystallography and NMR, PLA15 interaction energies) are external to the fitted free-energy labels and are not recovered by construction. The sole mild circularity is the presentation of the CombiSolv training residual (and partly-overlapping FreeSolv) as a “prediction” that demonstrates state-of-the-art accuracy. The acknowledged formal inconsistency of stage-2 force matching (vanishing Neff, classical configurations) is a correctness/bias risk, not a definitional loop. No self-definitional equations, uniqueness theorems, or load-bearing self-citations appear. Score remains low because the paper is self-contained against independent experimental and DFT-derived benchmarks.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 1 invented entities

The central claim rests on the MACE architecture, the SPICE/ωB97M-D3 reference level, the classical mdCATH sampling distribution, the ReSolv/DiffTRe free-energy path, and a collection of architectural and optimisation hyper-parameters. No new physical entities are postulated; the free parameters are the usual neural-network weights plus a handful of training-schedule choices.

free parameters (5)
  • MACE network weights (~1.31e6)
    All trainable parameters of the equivariant GNN; fitted sequentially to DFT, force-matching and free-energy losses.
  • force/energy loss weights (ωF, ωU)
    Hand-chosen relative weights that change between training stages (Methods).
  • cutoff radius 0.5 nm, max spherical harmonic order L=2
    Architectural hyper-parameters fixed by the authors to control cost and locality.
  • ReSolv effective-sample-size threshold N̄eff=0.9
    Controls when trajectories are regenerated during free-energy fine-tuning.
  • hydrogen-mass-repartitioning factor 2, Langevin friction 50/ps
    Simulation settings that affect the free-energy path used for top-down training.
axioms (4)
  • domain assumption Variational force matching on classical configurations yields a usable approximation to the true potential of mean force once free-energy reweighting is applied.
    Stated in Results (second training stage) and Methods; the paper acknowledges vanishing ESS and formal inconsistency.
  • domain assumption ωB97M-D3(BJ)/def2-TZVPPD DFT forces and energies are sufficiently accurate reference labels for biomolecular force fields.
    Inherited from the SPICE and MACE-OFF literature used to train TWIN-AT.
  • ad hoc to paper Local message-passing with 0.5 nm cutoff plus two-body ZBL repulsion is adequate for the many-body solvent-mediated interactions of interest.
    Architectural choice; the authors themselves note residual long-range deficiencies for NOEs and order parameters.
  • domain assumption Experimental hydration free energies in CombiSolv/FreeSolv are reliable top-down targets for continuum solvent effects.
    Standard assumption of the ReSolv framework.
invented entities (1)
  • TWIN (Transferable Water Implicit Network) no independent evidence
    purpose: Name for the final implicit-solvent MLP obtained after the three-stage training procedure.
    Not a new physical entity; simply the trained model. independent_evidence is false because the model is defined by the paper’s training pipeline.

pith-pipeline@v1.1.0-grok45 · 33112 in / 2992 out tokens · 26790 ms · 2026-07-14T08:31:08.349459+00:00 · methodology

0 comments
read the original abstract

Machine learning interatomic potentials (MLPs) have revolutionized atomistic modeling, offering the potential to replace traditional methods like Density Functional Theory (DFT). However, inference time of MLPs is orders of magnitude slower than that of classical force fields, hindering real-world applications for biomolecular systems that require timescales of microseconds and beyond. Implicit solvent MLPs can address this issue, but are faced with data challenges associated with coarse-grained modeling. Consequently, previous approaches relied on empirical force field data, thereby inherently limiting the MLP's accuracy. Here, we introduce the Transferable Water Implicit Network (TWIN), an implicit water MLP parametrized entirely by an Equivariant Graph Neural Network and trained solely on ab initio and experimental labels. We demonstrate TWIN's transferability across drug-like molecules, peptides, and proteins, achieving excellent results on ab initio and experimental crystallographic and NMR benchmarks, consistently outperforming previous machine-learning-based implicit solvent or coarse-grained models. Furthermore, TWIN closely matches DFT-based explicit solvent MLPs while providing a two-order-of-magnitude faster timestep evaluation, paving the way for efficient ab initio-level modeling of biomolecular systems in aqueous environments.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

119 extracted references · 63 canonical work pages

  1. [1]

    Nature Structural Biology 9(9), 646–652 (2002) https://doi.org/10.1038/nsb0902-646

    Karplus, M., McCammon, J.A.: Molecular dynamics simulations of biomolecules. Nature Structural Biology 9(9), 646–652 (2002) https://doi.org/10.1038/nsb0902-646

    work page doi:10.1038/nsb0902-646 2002

  2. [2]

    Annual Review of Biophysics41(Volume 41, 2012), 429–452 (2012) https: //doi.org/10.1146/annurev-biophys-042910-155245

    Dror, R.O., Dirks, R.M., Grossman, J.P., Xu, H., Shaw, D.E.: Biomolecular simulation: A computational microscope for molecular biology. Annual Review of Biophysics41(Volume 41, 2012), 429–452 (2012) https: //doi.org/10.1146/annurev-biophys-042910-155245

    doi:10.1146/annurev-biophys-042910-155245 2012

  3. [3]

    Neuron99(6), 1129–1143 (2018) https://doi.org/10.1016/j.neuron.2018.08.011

    Hollingsworth, S.A., Dror, R.O.: Molecular dynamics simulation for all. Neuron99(6), 1129–1143 (2018) https://doi.org/10.1016/j.neuron.2018.08.011

    doi:10.1016/j.neuron.2018.08.011 2018

  4. [4]

    BMC Biology9(1), 71 (2011) https://doi.org/10.1186/1741-7007-9-71

    Durrant, J.D., McCammon, J.A.: Molecular dynamics simulations and drug discovery. BMC Biology9(1), 71 (2011) https://doi.org/10.1186/1741-7007-9-71

    work page doi:10.1186/1741-7007-9-71 2011

  5. [5]

    Journal of Medicinal Chemistry59(9), 4035–4061 (2016) https://doi.org/10.1021/acs.jmedchem.5 b01684 https://doi.org/10.1021/acs.jmedchem.5b01684

    De Vivo, M., Masetti, M., Bottegoni, G., Cavalli, A.: Role of molecular dynamics and related methods in drug discovery. Journal of Medicinal Chemistry59(9), 4035–4061 (2016) https://doi.org/10.1021/acs.jmedchem.5 b01684 https://doi.org/10.1021/acs.jmedchem.5b01684. PMID: 26807648

    work page doi:10.1021/acs.jmedchem.5 2016

  6. [6]

    In: Protein Simulations

    Ponder, J.W., Case, D.A.: Force fields for protein simulations. In: Protein Simulations. Advances in Protein Chemistry, vol. 66, pp. 27–85. Academic Press, ??? (2003). https://doi.org/10.1016/S0065-3233(03)66002-X .https://www.sciencedirect.com/science/article/pii/S006532330366002X

    doi:10.1016/s0065-3233(03)66002-x 2003

  7. [7]

    Jour- nal of Computational Chemistry25(13), 1584–1604 (2004) https://doi.org/10.1002/jcc.20082 https://onlinelibrary.wiley.com/doi/pdf/10.1002/jcc.20082

    Mackerell Jr., A.D.: Empirical force fields for biological macromolecules: Overview and issues. Jour- nal of Computational Chemistry25(13), 1584–1604 (2004) https://doi.org/10.1002/jcc.20082 https://onlinelibrary.wiley.com/doi/pdf/10.1002/jcc.20082

    doi:10.1002/jcc.20082 2004

  8. [8]

    In: Kukol, A

    Lopes, P.E.M., Guvench, O., MacKerell, A.D.: Current Status of Protein Force Fields for Molecular Dynamics Simulations. In: Kukol, A. (ed.) Molecular Modeling of Proteins, pp. 47–71. Springer, New York, NY (2015). https://doi.org/10.1007/978-1-4939-1465-4 3

    work page doi:10.1007/978-1-4939-1465-4 2015

  9. [9]

    Current Opinion in Structural Biology67, 18–24 (2021) https://doi.org/10.1016/j.sbi.2020.08.006

    van der Spoel, D.: Systematic design of biomolecular force fields. Current Opinion in Structural Biology67, 18–24 (2021) https://doi.org/10.1016/j.sbi.2020.08.006 . Theory and Simulation/Computational Methods Macromolecular Assemblies

    work page doi:10.1016/j.sbi.2020.08.006 2021

  10. [10]

    https://arxiv.org/abs/2605.28960

    Kabylda, A., Esders, M., Gori, M., Chmiela, S., M¨ uller, K.-R., Tkatchenko, A.: How Atoms Interact Within Molecules (2026). https://arxiv.org/abs/2605.28960

    Pith/arXiv arXiv 2026

  11. [11]

    https://arxiv.org/abs/2603.24360

    St¨ ormer, I., Zavadlav, J.: Aluminum solidification and nanopolycrystal deformation via a Graph Neural Network Potential and Million-Atom Simulations (2026). https://arxiv.org/abs/2603.24360

    arXiv 2026

  12. [12]

    Journal of Chemical Theory and Computation21(15), 7550–7560 (2025) https://doi.org/10.1021/acs.jctc.5c00996 https://doi.org/10.1021/acs.jctc.5c00996

    Fuchs, P., Chen, W., Thaler, S., Zavadlav, J.: chemtrain-deploy: A parallel and scalable framework for machine learning potentials in million-atom md simulations. Journal of Chemical Theory and Computation21(15), 7550–7560 (2025) https://doi.org/10.1021/acs.jctc.5c00996 https://doi.org/10.1021/acs.jctc.5c00996. PMID: 40699940

    work page doi:10.1021/acs.jctc.5c00996 2025

  13. [13]

    In: Oh, A.H., Agarwal, A., Belgrave, D., Cho, K

    Batatia, I., Kovacs, D.P., Simm, G.N.C., Ortner, C., Csanyi, G.: MACE: Higher order equivariant message passing neural networks for fast and accurate force fields. In: Oh, A.H., Agarwal, A., Belgrave, D., Cho, K. (eds.) Advances in Neural Information Processing Systems (2022).https://openreview.net/forum?id=YPpSngE- ZU

    2022

  14. [14]

    https://arxiv.org/abs/2505.08762

    Levine, D.S., Shuaibi, M., Spotte-Smith, E.W.C., Taylor, M.G., Hasyim, M.R., Michel, K., Batatia, I., Cs´ anyi, G., Dzamba, M., Eastman, P., Frey, N.C., Fu, X., Gharakhanyan, V., Krishnapriyan, A.S., Rackers, J.A., Raja, S., Rizvi, A., Rosen, A.S., Ulissi, Z., Vargas, S., Zitnick, C.L., Blau, S.M., Wood, B.M.: The Open Molecules 2025 (OMol25) Dataset, Eva...

    arXiv 2025

  15. [15]

    https://arxiv.org/abs/2510.099 39 15

    Kabylda, A., Su´ arez-Dou, S., Davoine, N., Br¨ unig, F.N., Tkatchenko, A.: QCell: Comprehensive Quantum- Mechanical Dataset Spanning Diverse Biomolecular Fragments (2026). https://arxiv.org/abs/2510.099 39 15

    2026

  16. [16]

    Scientific Data10(1), 11 (2023) https://doi.org/10.1038/ s41597-022-01882-6

    Eastman, P., Behara, P.K., Dotson, D.L., Galvelis, R., Herr, J.E., Horton, J.T., Mao, Y., Chodera, J.D., Pritchard, B.P., Wang, Y., De Fabritiis, G., Markland, T.E.: SPICE, A Dataset of Drug-like Molecules and Peptides for Training Machine Learning Potentials. Scientific Data10(1), 11 (2023) https://doi.org/10.1038/ s41597-022-01882-6

    2023

  17. [17]

    Journal of the American Chemical Society147(21), 17598–17611 (2025) https://doi.org/10.1021/ jacs.4c07099 https://doi.org/10.1021/jacs.4c07099

    Kov´ acs, D.P., Moore, J.H., Browning, N.J., Batatia, I., Horton, J.T., Pu, Y., Kapil, V., Witt, W.C., Magd˘ au, I.-B., Cole, D.J., Cs´ anyi, G.: Mace-off: Short-range transferable machine learning force fields for organic molecules. Journal of the American Chemical Society147(21), 17598–17611 (2025) https://doi.org/10.1021/ jacs.4c07099 https://doi.org/1...

    doi:10.1021/jacs.4c07099 2025

  18. [18]

    Journal of the American Chemical Society147(37), 33723–33734 (2025) https://doi.or g/10.1021/jacs.5c09558 https://doi.org/10.1021/jacs.5c09558

    Kabylda, A., Frank, J.T., Su´ arez-Dou, S., Khabibrakhmanov, A., Medrano Sandonas, L., Unke, O.T., Chmiela, S., M¨ uller, K.-R., Tkatchenko, A.: Molecular simulations with a pretrained neural network and universal pairwise force fields. Journal of the American Chemical Society147(37), 33723–33734 (2025) https://doi.or g/10.1021/jacs.5c09558 https://doi.or...

    doi:10.1021/jacs.5c09558 2025

  19. [19]

    Science Advances10(14), 4397 (2024) https://doi.org/10.1126/sciadv.adn4397 https://www.science.org/doi/pdf/10.1126/sciadv.adn4397

    Unke, O.T., St¨ ohr, M., Ganscha, S., Unterthiner, T., Maennel, H., Kashubin, S., Ahlin, D., Gastegger, M., Sandonas, L.M., Berryman, J.T., Tkatchenko, A., M¨ uller, K.-R.: Biomolecular dynamics with machine-learned quantum-mechanical force fields trained on diverse chemical fragments. Science Advances10(14), 4397 (2024) https://doi.org/10.1126/sciadv.adn...

    doi:10.1126/sciadv.adn4397 2024

  20. [20]

    Living Journal of Computational Molecular Science4(1), 1583 (2022) https://doi.org/10.33011/livecoms.4.1.1583

    H´ enin, J., Leli` evre, T., Shirts, M.R., Valsson, O., Delemotte, L.: Enhanced Sampling Methods for Molecular Dynamics Simulations [Article v1.0]. Living Journal of Computational Molecular Science4(1), 1583 (2022) https://doi.org/10.33011/livecoms.4.1.1583 . Chap. Articles. Accessed 2026-05-07

    doi:10.33011/livecoms.4.1.1583 2022

  21. [21]

    Annual Review of Physical Chemistry75(Volume 75, 2024), 347–370 (2024) https://doi.org/10.1146/annurev-physc hem-083122-125941

    Mehdi, S., Smith, Z., Herron, L., Zou, Z., Tiwary, P.: Enhanced sampling with machine learning. Annual Review of Physical Chemistry75(Volume 75, 2024), 347–370 (2024) https://doi.org/10.1146/annurev-physc hem-083122-125941

    work page doi:10.1146/annurev-physc 2024

  22. [22]

    Chemical Reviews126(1), 671–713 (2026) https://doi.org/10.1021/ac s.chemrev.5c00700 https://doi.org/10.1021/acs.chemrev.5c00700

    Zhu, K., Trizio, E., Zhang, J., Hu, R., Jiang, L., Hou, T., Bonati, L.: Enhanced sampling in the age of machine learning: Algorithms and applications. Chemical Reviews126(1), 671–713 (2026) https://doi.org/10.1021/ac s.chemrev.5c00700 https://doi.org/10.1021/acs.chemrev.5c00700. PMID: 41124671

    work page doi:10.1021/ac 2026

  23. [24]

    International Journal of Molecular Sciences20(15), 3774 (2019) https://doi.org/10.3390/ijms20153774

    Singh, N., Li, W.: Recent Advances in Coarse-Grained Models for Biomolecules and Their Applications. International Journal of Molecular Sciences20(15), 3774 (2019) https://doi.org/10.3390/ijms20153774

    work page doi:10.3390/ijms20153774 2019

  24. [25]

    ACS Central Science9(12), 2286–2297 (2023) https://doi.org/10.1021/acscentsci.3c01160 https://doi.org/10.1021/acscentsci.3c01160

    Airas, J., Ding, X., Zhang, B.: Transferable implicit solvation via contrastive learning of graph neural networks. ACS Central Science9(12), 2286–2297 (2023) https://doi.org/10.1021/acscentsci.3c01160 https://doi.org/10.1021/acscentsci.3c01160

    work page doi:10.1021/acscentsci.3c01160 2023

  25. [26]

    Journal of Chemical The- ory and Computation18(10), 6334–6344 (2022) ht t p s : / / d o i

    Ding, X., Zhang, B.: Contrastive learning of coarse-grained force fields. Journal of Chemical The- ory and Computation18(10), 6334–6344 (2022) ht t p s : / / d o i . o r g / 1 0 . 1 0 2 1 / a c s . j c t c . 2 c 0 0 6 16 https://doi.org/10.1021/acs.jctc.2c00616. PMID: 36112935

    work page doi:10.1021/acs.jctc.2c00616 2022

  26. [27]

    Annual Review of Biophysics48(Volume 48, 2019), 275–296 (2019) https://doi.org/10.1146/annurev-biophys-052118-115325

    Onufriev, A.V., Case, D.A.: Generalized born implicit solvent models for biomolecules. Annual Review of Biophysics48(Volume 48, 2019), 275–296 (2019) https://doi.org/10.1146/annurev-biophys-052118-115325

    doi:10.1146/annurev-biophys-052118-115325 2019

  27. [28]

    The Journal of Chemical Physics155(8), 084101 (2021) https://doi.org/10.1063/5

    Chen, Y., Kr¨ amer, A., Charron, N.E., Husic, B.E., Clementi, C., No´ e, F.: Machine learning implicit solvation for molecular dynamics. The Journal of Chemical Physics155(8), 084101 (2021) https://doi.org/10.1063/5. 0059915 . Accessed 2026-07-05

    doi:10.1063/5 2021

  28. [29]

    Nature Communications14(1), 5739 (2023) https://doi.org/10.1038/s41467-023-41343-1

    Majewski, M., P´ erez, A., Th¨ olke, P., Doerr, S., Charron, N.E., Giorgino, T., Husic, B.E., Clementi, C., No´ e, F., De Fabritiis, G.: Machine learning coarse-grained potentials of protein thermodynamics. Nature Communications14(1), 5739 (2023) https://doi.org/10.1038/s41467-023-41343-1

    work page doi:10.1038/s41467-023-41343-1 2023

  29. [30]

    ACS Central Science5(5), 755–767 (2019) https: //doi.org/10.1021/acscentsci.8b00913 https://doi.org/10.1021/acscentsci.8b00913

    Wang, J., Olsson, S., Wehmeyer, C., P´ erez, A., Charron, N.E., Fabritiis, G., No´ e, F., Clementi, C.: Machine 16 learning of coarse-grained molecular dynamics force fields. ACS Central Science5(5), 755–767 (2019) https: //doi.org/10.1021/acscentsci.8b00913 https://doi.org/10.1021/acscentsci.8b00913. PMID: 31139712

    doi:10.1021/acscentsci.8b00913 2019

  30. [31]

    Annual Review of Physical Chemistry75(Volume 75, 2024), 21–45 (2024) https://doi.org/10.1146/annurev-physche m-062123-010821

    Noid, W.G., Szukalo, R.J., Kidder, K.M., Lesniewski, M.C.: Rigorous progress in coarse-graining. Annual Review of Physical Chemistry75(Volume 75, 2024), 21–45 (2024) https://doi.org/10.1146/annurev-physche m-062123-010821

    work page doi:10.1146/annurev-physche 2024

  31. [32]

    The Journal of Chemical Physics164(5) (2026) https://doi.org/10.1063/5.0313624

    Slejko, E., Coste, A., Potisk, T., Zavadlav, J., Praprotnik, M.: Achieving all-atom molecular dynamics accuracy from the poisson–boltzmann method through machine learning. The Journal of Chemical Physics164(5) (2026) https://doi.org/10.1063/5.0313624

    work page doi:10.1063/5.0313624 2026

  32. [33]

    Journal of Chemical Theory and Computation (2023) https://doi

    Coste, A., Slejko, E., Zavadlav, J., Praprotnik, M.: Developing an implicit solvation machine learning model for molecular simulations of ionic media. Journal of Chemical Theory and Computation (2023) https://doi. org/10.1021/acs.jctc.3c00984

    doi:10.1021/acs.jctc.3c00984 2023

  33. [34]

    The Journal of Chemical Physics157(24), 244103 (2022) https://doi.org/10.1063/5.0124538

    Thaler, S., Stupp, M., Zavadlav, J.: Deep coarse-grained potentials via relative entropy minimization. The Journal of Chemical Physics157(24), 244103 (2022) https://doi.org/10.1063/5.0124538 . Accessed 2026-07-04

    doi:10.1063/5.0124538 2022

  34. [35]

    Katzberger, P., Riniker, S.: A general graph neural network based implicit solvation model for organic molecules in water. Chem. Sci.15, 10794–10802 (2024) https://doi.org/10.1039/D4SC02432J

    work page doi:10.1039/d4sc02432j 2024

  35. [36]

    Nature Chemistry (2025) https://doi.org/10.1038/s415 57-025-01874-0

    Charron, N.E., Bonneau, K., Pasos-Trejo, A.S., Guljas, A., Chen, Y., Musil, F., Venturin, J., Gusew, D., Zaporozhets, I., Kr¨ amer, A., Templeton, C., Kelkar, A., Durumeric, A.E.P., Olsson, S., P´ erez, A., Majewski, M., Husic, B.E., Patel, A., De Fabritiis, G., No´ e, F., Clementi, C.: Navigating protein landscapes with a machine-learned transferable coa...

    doi:10.1038/s415 2025

  36. [37]

    Nature Communications17(1), 2493 (2026) https://doi.org/10.1 038/s41467-026-70818-0

    Durumeric, A.E.P., Chen, Y., Pasos-Trejo, A.S., No´ e, F., Clementi, C.: Learning data-efficient coarse-grained molecular dynamics from forces and noise. Nature Communications17(1), 2493 (2026) https://doi.org/10.1 038/s41467-026-70818-0

    2026

  37. [38]

    Journal of Chemical Theory and Computation22(1), 219–230 (2026) https: //doi.org/10.1021/acs.jctc.5c01712 https://doi.org/10.1021/acs.jctc.5c01712

    Chen, W., G¨ orlich, F., Fuchs, P., Zavadlav, J.: Enhanced sampling for efficient learning of coarse-grained machine learning potentials. Journal of Chemical Theory and Computation22(1), 219–230 (2026) https: //doi.org/10.1021/acs.jctc.5c01712 https://doi.org/10.1021/acs.jctc.5c01712. PMID: 41437682

    work page doi:10.1021/acs.jctc.5c01712 2026

  38. [39]

    Journal of chemical theory and computation18(10), 6334–6344 (2022)

    Ding, X., Zhang, B.: Contrastive learning of coarse-grained force fields. Journal of chemical theory and computation18(10), 6334–6344 (2022)

    2022

  39. [40]

    ACS Central Science9(12), 2286–2297 (2023)

    Airas, J., Ding, X., Zhang, B.: Transferable implicit solvation via contrastive learning of graph neural networks. ACS Central Science9(12), 2286–2297 (2023)

    2023

  40. [41]

    The Journal of Physical Chemistry99(7), 2224–2235 (1995)

    Klamt, A.: Conductor-like screening model for real solvents: a new approach to the quantitative calculation of solvation phenomena. The Journal of Physical Chemistry99(7), 2224–2235 (1995)

    1995

  41. [42]

    Journal of Chemical Theory and Computation9(4), 2052–2071 (2013) https://doi.org/10.1021/ct301050x https://doi.org/10.1021/ct301050x

    Barone, V., Carnimeo, I., Scalmani, G.: Computational spectroscopy of large systems in solution: The dft- b/pcm and td-dftb/pcm approach. Journal of Chemical Theory and Computation9(4), 2052–2071 (2013) https://doi.org/10.1021/ct301050x https://doi.org/10.1021/ct301050x. PMID: 26583552

    work page doi:10.1021/ct301050x 2052

  42. [43]

    Jour- nal of Chemical Theory and Computation6(8), 2303–2314 (2010) https://doi.org/10.1021/ct1001818 https://doi.org/10.1021/ct1001818

    Hou, G., Zhu, X., Cui, Q.: An implicit solvent model for scc-dftb with charge-dependent radii. Jour- nal of Chemical Theory and Computation6(8), 2303–2314 (2010) https://doi.org/10.1021/ct1001818 https://doi.org/10.1021/ct1001818. PMID: 20711513

    work page doi:10.1021/ct1001818 2010

  43. [44]

    The Journal of Chemical Physics161(23), 234101 (2024) https://doi.org/10.1063/5.0235 189

    R¨ ocken, S., Burnet, A.F., Zavadlav, J.: Predicting solvation free energies with an implicit solvent machine learning potential. The Journal of Chemical Physics161(23), 234101 (2024) https://doi.org/10.1063/5.0235 189

    work page doi:10.1063/5.0235 2024

  44. [45]

    In: Oh, A.H., Agarwal, A., Belgrave, D., Cho, K

    Batatia, I., Kovacs, D.P., Simm, G.N.C., Ortner, C., Csanyi, G.: MACE: Higher order equivariant message passing neural networks for fast and accurate force fields. In: Oh, A.H., Agarwal, A., Belgrave, D., Cho, K. 17 (eds.) Advances in Neural Information Processing Systems (2022).https://openreview.net/forum?id=YPpSngE- ZU

    2022

  45. [46]

    Journal of Chemical Theory and Computation20(19), 8583–8593 (2024) https://doi.org/10.1021/acs.jctc.4c00794 https://doi.org/10.1021/acs.jctc.4c00794

    Eastman, P., Pritchard, B.P., Chodera, J.D., Markland, T.E.: Nutmeg and spice: Models and data for biomolecular machine learning. Journal of Chemical Theory and Computation20(19), 8583–8593 (2024) https://doi.org/10.1021/acs.jctc.4c00794 https://doi.org/10.1021/acs.jctc.4c00794. PMID: 39318326

    doi:10.1021/acs.jctc.4c00794 2024

  46. [48]

    Mobley, D.L., Guthrie, J.P.: Freesolv: a database of experimental and calculated hydration free energies, with input files. J. Comput. Aided Mol. Des.28(7), 711–720 (2014)

    2014

  47. [49]

    The Journal of Physical Chemistry B128(28), 6693–6703 (2024) https: //doi.org/10.1021/acs.jpcb.4c01417 https://doi.org/10.1021/acs.jpcb.4c01417

    Karwounopoulos, J., Wu, Z., Tkaczyk, S., Wang, S., Baskerville, A., Ranasinghe, K., Langer, T., Wood, G.P.F., Wieder, M., Boresch, S.: Insights and challenges in correcting force field based solvation free energies using a neural network potential. The Journal of Physical Chemistry B128(28), 6693–6703 (2024) https: //doi.org/10.1021/acs.jpcb.4c01417 https...

    work page doi:10.1021/acs.jpcb.4c01417 2024

  48. [50]

    Journal of the American Chemical Society148(5), 4928–4937 (2026) https://doi.org/10.1 021/jacs.5c10940 https://doi.org/10.1021/jacs.5c10940

    Harry Moore, J., Cole, D.J., Cs´ anyi, G.: Computing solvation free energies of small molecules with experi- mental accuracy. Journal of the American Chemical Society148(5), 4928–4937 (2026) https://doi.org/10.1 021/jacs.5c10940 https://doi.org/10.1021/jacs.5c10940. PMID: 41591329

    work page doi:10.1021/jacs.5c10940 2026

  49. [51]

    Scientific Data10(1), 11 (2023) https://doi.org/10.1038/ s41597-022-01882-6

    Eastman, P., Behara, P.K., Dotson, D.L., Galvelis, R., Herr, J.E., Horton, J.T., Mao, Y., Chodera, J.D., Pritchard, B.P., Wang, Y., De Fabritiis, G., Markland, T.E.: SPICE, A Dataset of Drug-like Molecules and Peptides for Training Machine Learning Potentials. Scientific Data10(1), 11 (2023) https://doi.org/10.1038/ s41597-022-01882-6 . Publisher: Nature ...

    2023

  50. [52]

    Journal of Chemical Theory and Computation16(7), 4192–4202 (2020) https://doi.org/10.1021/acs.jctc.0c00121 https://doi.org/10.1021/acs.jctc.0c00121

    Devereux, C., Smith, J.S., Huddleston, K.K., Barros, K., Zubatyuk, R., Isayev, O., Roitberg, A.E.: Extend- ing the applicability of the ani deep learning molecular potential to sulfur and halogens. Journal of Chemical Theory and Computation16(7), 4192–4202 (2020) https://doi.org/10.1021/acs.jctc.0c00121 https://doi.org/10.1021/acs.jctc.0c00121. PMID: 32543858

    work page doi:10.1021/acs.jctc.0c00121 2020

  51. [53]

    first-principles

    Medders, G.R., Babin, V., Paesani, F.: Development of a “first-principles” water potential with flexible monomers. iii. liquid phase properties. Journal of Chemical Theory and Computation10(8), 2906–2910 (2014) https://doi.org/10.1021/ct5004115 https://doi.org/10.1021/ct5004115. PMID: 26588266

    work page doi:10.1021/ct5004115 2014

  52. [54]

    Chemical Physics258(2), 121–137 (2000) https://doi.org/10.1016/S0301-0104(00)00179-8

    Soper, A.K.: The radial distribution functions of water and ice from 220 to 673 k and at pressures up to 400 mpa. Chemical Physics258(2), 121–137 (2000) https://doi.org/10.1016/S0301-0104(00)00179-8

    work page doi:10.1016/s0301-0104(00)00179-8 2000

  53. [55]

    Nucleic Acids Research43(D1), 376–381 (2014) https://doi.or g/10.1093/nar/gku947 https://academic.oup.com/nar/article-pdf/43/D1/D376/7330586/gku947.pdf

    Sillitoe, I., Lewis, T.E., Cuff, A., Das, S., Ashford, P., Dawson, N.L., Furnham, N., Laskowski, R.A., Lee, D., Lees, J.G., Lehtinen, S., Studer, R.A., Thornton, J., Orengo, C.A.: Cath: comprehensive structural and functional annotations for genome sequences. Nucleic Acids Research43(D1), 376–381 (2014) https://doi.or g/10.1093/nar/gku947 https://academic...

    work page doi:10.1093/nar/gku947 2014

  54. [56]

    https://arxiv.org/abs/2407.14794

    Mirarchi, A., Giorgino, T., Fabritiis, G.D.: mdCATH: A Large-Scale MD Dataset for Data-Driven Computa- tional Biophysics (2024). https://arxiv.org/abs/2407.14794

    Pith/arXiv arXiv 2024

  55. [57]

    Chemical Engineering Journal418, 129307 (2021) https://doi.org/10.1016/j.cej.2021.129307

    Vermeire, F.H., Green, W.H.: Transfer learning for solvation free energies: From quantum chemistry to experiments. Chemical Engineering Journal418, 129307 (2021) https://doi.org/10.1016/j.cej.2021.129307

    doi:10.1016/j.cej.2021.129307 2021

  56. [58]

    Nature Communications12(1), 6884 (2021) https://doi.org/10.1038/s41467-021-27241-4

    Thaler, S., Zavadlav, J.: Learning neural network potentials from experimental data via Differentiable Trajec- tory Reweighting. Nature Communications12(1), 6884 (2021) https://doi.org/10.1038/s41467-021-27241-4 . Publisher: Nature Publishing Group. Accessed 2024-03-05

    doi:10.1038/s41467-021-27241-4 2021

  57. [59]

    Zwanzig, R.W.: High-temperature equation of state by a perturbation method. i. nonpolar gases. The Journal of Chemical Physics22(8), 1420–1426 (1954) https://doi.org/10.1063/1.1740409 18

    work page doi:10.1063/1.1740409 1954

  58. [61]

    Journal of Chemical Information and Modeling60(12), 6258–6268 (2020) https://doi.org/10.1021/acs.jcim.0c00904 https://doi.org/10.1021/acs.jcim.0c00904

    Lahey, S.-L.J., Thien Phuc, T.N., Rowley, C.N.: Benchmarking force field and the ani neural net- work potentials for the torsional potential energy surface of biaryl drug fragments. Journal of Chemical Information and Modeling60(12), 6258–6268 (2020) https://doi.org/10.1021/acs.jcim.0c00904 https://doi.org/10.1021/acs.jcim.0c00904. PMID: 33263401

    work page doi:10.1021/acs.jcim.0c00904 2020

  59. [62]

    Journal of Chemical Information and Modeling64(20), 7938–7948 (2024) https://doi.org/10.1021/acs.jcim.4c01120 https://doi.org/10.1021/acs.jcim.4c01120

    Waibl, F., Casagrande, F., Dey, F., Riniker, S.: Validating small-molecule force fields for macrocyclic com- pounds using nmr data in different solvents. Journal of Chemical Information and Modeling64(20), 7938–7948 (2024) https://doi.org/10.1021/acs.jcim.4c01120 https://doi.org/10.1021/acs.jcim.4c01120. PMID: 39405498

    work page doi:10.1021/acs.jcim.4c01120 2024

  60. [63]

    The Lancet Neurology15(4), 373–381 (2016) https://doi.org/10.1016/S1474-4422(16)00018-1

    Cohen, J.A., Arnold, D.L., Comi, G., Bar-Or, A., Gujrathi, S., Hartung, J.P., Cravets, M., Olson, A., Frohna, P.A., Selmaj, K.W.: Safety and efficacy of the selective sphingosine 1-phosphate receptor modulator ozani- mod in relapsing multiple sclerosis (radiance): a randomised, placebo-controlled, phase 2 trial. The Lancet Neurology15(4), 373–381 (2016) h...

    work page doi:10.1016/s1474-4422(16)00018-1 2016

  61. [64]

    Nature Reviews Drug Discovery7(7), 608–624 (2008) https://doi.org/10.103 8/nrd2590

    Driggers, E.M., Hale, S.P., Lee, J., Terrett, N.K.: The exploration of macrocycles for drug discovery — an underexploited structural class. Nature Reviews Drug Discovery7(7), 608–624 (2008) https://doi.org/10.103 8/nrd2590 . Accessed 2026-02-25

    2008

  62. [65]

    Journal of Chemical Information and Modeling63(1), 138–146 (2023) https://doi.org/10.1021/acs.jcim.2c01093 https://doi.org/10.1021/acs.jcim.2c01093

    Sethio, D., Poongavanam, V., Xiong, R., Tyagi, M., Duy Vo, D., Lindh, R., Kihlberg, J.: Simulation reveals the chameleonic behavior of macrocycles. Journal of Chemical Information and Modeling63(1), 138–146 (2023) https://doi.org/10.1021/acs.jcim.2c01093 https://doi.org/10.1021/acs.jcim.2c01093. PMID: 36563083

    work page doi:10.1021/acs.jcim.2c01093 2023

  63. [66]

    Journal of Chemical Information and Modeling58(5), 982–992 (2018) https://doi.org/10.1021/acs.jcim.8b00097 https://doi.org/10.1021/acs.jcim.8b00097

    Kamenik, A.S., Lessel, U., Fuchs, J.E., Fox, T., Liedl, K.R.: Peptidic macrocycles - conformational sampling and thermodynamic characterization. Journal of Chemical Information and Modeling58(5), 982–992 (2018) https://doi.org/10.1021/acs.jcim.8b00097 https://doi.org/10.1021/acs.jcim.8b00097. PMID: 29652495

    work page doi:10.1021/acs.jcim.8b00097 2018

  64. [67]

    PMID: 31751129

    Zhang, S., Schweitzer-Stenner, R., Urbanc, B.: Do molecular dynamics force fields capture conformational dynamics of alanine in water? Journal of Chemical Theory and Computation16(1), 510–527 (2020) https: //doi.org/10.1021/acs.jctc.9b00588 https://doi.org/10.1021/acs.jctc.9b00588. PMID: 31751129

    work page doi:10.1021/acs.jctc.9b00588 2020

  65. [69]

    The Journal of Physical Chemistry B125(14), 3598–3612 (2021) https://doi.org/ 10.1021/acs.jpcb.0c10401 https://doi.org/10.1021/acs.jpcb.0c10401

    Rosenberger, D., Smith, J.S., Garcia, A.E.: Modeling of peptides with classical and novel machine learning force fields: A comparison. The Journal of Physical Chemistry B125(14), 3598–3612 (2021) https://doi.org/ 10.1021/acs.jpcb.0c10401 https://doi.org/10.1021/acs.jpcb.0c10401. PMID: 33798336

    work page doi:10.1021/acs.jpcb.0c10401 2021

  66. [70]

    The Journal of Physical Chemistry Letters12(32), 7701–7707 (2021) https://doi.org/10.1021/acs.jpclett.1c01987 https://doi.org/10.1021/acs.jpclett.1c01987

    Engel, E.A., Kapil, V., Ceriotti, M.: Importance of nuclear quantum effects for nmr crystallography. The Journal of Physical Chemistry Letters12(32), 7701–7707 (2021) https://doi.org/10.1021/acs.jpclett.1c01987 https://doi.org/10.1021/acs.jpclett.1c01987. PMID: 34355903

    work page doi:10.1021/acs.jpclett.1c01987 2021

  67. [71]

    Journal of Computational Chemistry34(25), 2135–2145 (2013) https://doi.org/10.1002/jc c.23354 https://onlinelibrary.wiley.com/doi/pdf/10.1002/jcc.23354

    Huang, J., MacKerell Jr, A.D.: Charmm36 all-atom additive protein force field: Validation based on compar- ison to nmr data. Journal of Computational Chemistry34(25), 2135–2145 (2013) https://doi.org/10.1002/jc c.23354 https://onlinelibrary.wiley.com/doi/pdf/10.1002/jcc.23354

    work page doi:10.1002/jc 2013

  68. [72]

    Biophysical Journal99(2), 647–655 (2010) https://doi.org/10.1016/j.bp j.2010.04.062

    Lange, O.F., van der Spoel, D., de Groot, B.L.: Scrutinizing molecular mechanics force fields on the submi- crosecond timescale with nmr data. Biophysical Journal99(2), 647–655 (2010) https://doi.org/10.1016/j.bp j.2010.04.062

    work page doi:10.1016/j.bp 2010

  69. [73]

    PLOS ONE7(2), 1–6 (2012) https://doi.org/10.1371/jour nal.pone.0032131

    Lindorff-Larsen, K., Maragakis, P., Piana, S., Eastwood, M.P., Dror, R.O., Shaw, D.E.: Systematic validation of protein force fields against experimental data. PLOS ONE7(2), 1–6 (2012) https://doi.org/10.1371/jour nal.pone.0032131

    doi:10.1371/jour 2012

  70. [75]

    Journal of the American Chemical Society124(47), 14221–14226 (2002) https://doi.org/10.1021/ja0273288 https://doi.org/10.1021/ja0273288

    Jurani´ c, N., Moncrieffe, M.C., Liki´ c, V.A., Prendergast, F.G., Macura, S.: Structural dependencies of h3jnc‘ scalar coupling in protein h-bond chains. Journal of the American Chemical Society124(47), 14221–14226 (2002) https://doi.org/10.1021/ja0273288 https://doi.org/10.1021/ja0273288. PMID: 12440921

    work page doi:10.1021/ja0273288 2002

  71. [76]

    Protein Science10(9), 1856–1868 (2001) https://doi.org/10.1110/ps.14301 https://onlinelibrary.wiley.com/doi/pdf/10.1110/ps.14301

    Alexandrescu, A.T., Snyder, D.R., Abildgaard, F.: Nmr of hydrogen bonding in cold-shock pro- tein a and an analysis of the influence of crystallographic resolution on comparisons of hydro- gen bond lengths. Protein Science10(9), 1856–1868 (2001) https://doi.org/10.1110/ps.14301 https://onlinelibrary.wiley.com/doi/pdf/10.1110/ps.14301

    work page doi:10.1110/ps.14301 2001

  72. [78]

    Journal of the American Chemical Society117(50), 12562–12566 (1995) https://doi.org/10 .1021/ja00155a020 https://doi.org/10.1021/ja00155a020

    Tjandra, N., Feller, S.E., Pastor, R.W., Bax, A.: Rotational diffusion anisotropy of human ubiquitin from 15n nmr relaxation. Journal of the American Chemical Society117(50), 12562–12566 (1995) https://doi.org/10 .1021/ja00155a020 https://doi.org/10.1021/ja00155a020

    work page doi:10.1021/ja00155a020 1995

  73. [79]

    Journal of the American Chemical Society133(4), 909–919 (2011) https://doi.org/10.1021/ja107847d https://doi.org/10.1021/ja107847d

    Esadze, A., Li, D.-W., Wang, T., Br¨ uschweiler, R., Iwahara, J.: Dynamics of lysine side-chain amino groups in a protein studied by heteronuclear 1h-15n nmr spectroscopy. Journal of the American Chemical Society133(4), 909–919 (2011) https://doi.org/10.1021/ja107847d https://doi.org/10.1021/ja107847d. PMID: 21186799

    work page doi:10.1021/ja107847d 2011

  74. [80]

    Journal of the American Chemical Society121(12), 2891–2902 (1999) https://doi.org/10.1021/ja983758f https://doi.org/10.1021/ja983758f

    Lee, A.L., Flynn, P.F., Wand, A.J.: Comparison of 2h and 13c nmr relaxation techniques for the study of protein methyl group dynamics in solution. Journal of the American Chemical Society121(12), 2891–2902 (1999) https://doi.org/10.1021/ja983758f https://doi.org/10.1021/ja983758f

    work page doi:10.1021/ja983758f 1999

  75. [81]

    Journal of Chemi- cal Theory and Computation21(24), 12709–12724 (2025) https://doi.org/10.1021/acs.jctc.5c01400 https://doi.org/10.1021/acs.jctc.5c01400

    Kim, D., Wang, X., Vargas, S., Zhong, P., King, D.S., Inizan, T.J., Cheng, B.: A universal augmentation framework for long-range electrostatics in machine learning interatomic potentials. Journal of Chemi- cal Theory and Computation21(24), 12709–12724 (2025) https://doi.org/10.1021/acs.jctc.5c01400 https://doi.org/10.1021/acs.jctc.5c01400. PMID: 41368735

    doi:10.1021/acs.jctc.5c01400 2025

  76. [82]

    Journal of Chemical Theory and Computation11(7), 3420–3431 (2015) https://doi.org/10.1021/ct501178z https://doi.org/10.1021/ct501178z

    Henriques, J., Cragnell, C., Skep¨ o, M.: Molecular dynamics simulations of intrinsically disordered proteins: Force field evaluation and comparison with experiment. Journal of Chemical Theory and Computation11(7), 3420–3431 (2015) https://doi.org/10.1021/ct501178z https://doi.org/10.1021/ct501178z. PMID: 26575776

    work page doi:10.1021/ct501178z 2015

  77. [83]

    Current Opinion in Structural Biology48, 40–48 (2018) https://doi.org/10.1016/j.sbi.2017.10.008

    Huang, J., MacKerell, A.D.: Force field development and simulations of intrinsically disordered proteins. Current Opinion in Structural Biology48, 40–48 (2018) https://doi.org/10.1016/j.sbi.2017.10.008 . Folding and binding in silico, in vitro and in cellula•Proteins: An Evolutionary Perspective

    work page doi:10.1016/j.sbi.2017.10.008 2018

  78. [84]

    Journal of Chemical Theory and Computation11(11), 5513–5524 (2015) https://doi.org/10.1021/acs.jctc.5b00736 https://doi.org/10.1021/acs.jctc.5b00736

    Rauscher, S., Gapsys, V., Gajda, M.J., Zweckstetter, M., Groot, B.L., Grubm¨ uller, H.: Structural ensembles of intrinsically disordered proteins depend strongly on force field: A comparison to experiment. Journal of Chemical Theory and Computation11(11), 5513–5524 (2015) https://doi.org/10.1021/acs.jctc.5b00736 https://doi.org/10.1021/acs.jctc.5b00736. P...

    work page doi:10.1021/acs.jctc.5b00736 2015

  79. [85]

    The Journal of Chemical Physics153(19), 194101 (2020) https://doi.org/10.1063/5.0026133

    Husic, B.E., Charron, N.E., Lemm, D., Wang, J., P´ erez, A., Majewski, M., Kr¨ amer, A., Chen, Y., Olsson, S., Fabritiis, G., No´ e, F., Clementi, C.: Coarse graining molecular dynamics with graph neural networks. The Journal of Chemical Physics153(19), 194101 (2020) https://doi.org/10.1063/5.0026133

    doi:10.1063/5.0026133 2020

  80. [86]

    The Journal of Chemical Physics 139(9), 090901 (2013) https://doi.org/10.1063/1.4818908

    Noid, W.G.: Perspective: Coarse-grained models for biomolecular systems. The Journal of Chemical Physics 139(9), 090901 (2013) https://doi.org/10.1063/1.4818908

    doi:10.1063/1.4818908 2013

Showing first 80 references.