pith. sign in

arxiv: 2605.17753 · v1 · pith:3DFXZYXXnew · submitted 2026-05-18 · ⚛️ physics.chem-ph

Fast and accurate committor estimation for kinetics simulations

Ru Wang , Xiaojun Ji , Hao Wang , Wenjian Liu This is my paper

Pith reviewed 2026-05-20 01:25 UTC · model grok-4.3

classification ⚛️ physics.chem-ph
keywords committor estimationMilestoningmean first passage timeneural networkbiomolecular kineticsshort trajectory simulationsanalogue prediction
0
0 comments X

The pith

A neural-network committor from short parallel trajectories enables low-cost Milestoning to compute mean first passage times for biomolecular kinetics.

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

The paper develops a fast algorithm for estimating the committor, the optimal reaction coordinate for biomolecular kinetics. It uses highly parallelizable short trajectory simulations combined with analogue prediction to train a neural network representation of the committor. This committor is then integrated with the Milestoning method to predict the mean first passage time at very low computational cost. The approach is demonstrated to be robust across systems of increasing complexity.

Core claim

The committor-guided Milestoning (CoM) method obtains an accurate neural network committor from short trajectory simulations and analogue prediction, then couples it with Milestoning to predict the mean first passage time at very low computational cost for biomolecular processes.

What carries the argument

The committor-guided Milestoning (CoM) method, which represents the committor via a neural network ansatz derived from short trajectories and analogue prediction to guide Milestoning for MFPT prediction.

Load-bearing premise

Highly parallelizable short trajectory simulations combined with analogue prediction yield a sufficiently accurate neural-network representation of the committor for the Milestoning step to produce reliable MFPT values in systems of increasing complexity.

What would settle it

Running the CoM method on a system with a known exact mean first passage time from long simulations or analytical solution and finding significant discrepancy would falsify the accuracy claim.

read the original abstract

Computing long-timescale kinetics of biomolecular processes remains a major challenge for atomistic simulations. A way out is to exploit local kinetic information to construct the global stationary flux across the reaction space. The committor serves as the optimal reaction coordinate for this purpose; however, its calculation is itself highly demanding. Here, we introduce a fast and accurate algorithm for committor estimation by leveraging highly parallelizable short trajectory simulations and analogue prediction. The resulting committor is represented via a neural network ansatz and subsequently coupled with the Milestoning method to predict the mean first passage time at very low computational cost. We demonstrate the robustness and efficiency of this committor-guided Milestoning (CoM) method through examples of increasing complexity.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces the committor-guided Milestoning (CoM) method for computing long-timescale kinetics in biomolecular systems. It estimates the committor via a neural-network ansatz trained on short, highly parallelizable trajectories combined with analogue prediction, then couples this committor with Milestoning to obtain mean first passage times (MFPT) at low cost. Robustness is demonstrated through examples of increasing complexity.

Significance. If the accuracy claims hold, the approach could meaningfully reduce the computational burden of kinetics calculations by exploiting parallel short simulations and a learned reaction coordinate within an established Milestoning framework. The emphasis on parallelizability and the neural-network representation of the committor are clear strengths for scalability.

major comments (2)
  1. [Results section] Results section (examples of increasing complexity): the manuscript asserts robustness and accuracy but provides no quantitative propagation analysis showing how local committor errors—especially near the dividing surface—translate into MFPT bias in the Milestoning flux integrals. Short trajectories may undersample transition-state regions in high-barrier cases, and this risk is not explicitly bounded.
  2. [Methods section] Methods section (analogue prediction step): the procedure for extending committor values from short trajectories is described at a high level, yet no error metrics, bias controls, or sensitivity tests are reported that would confirm the neural-network surface remains sufficiently accurate for subsequent Milestoning.
minor comments (1)
  1. [Abstract] Abstract: quantitative statements on wall-clock savings or error reduction relative to standard committor methods would strengthen the efficiency claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments and positive assessment of the potential impact of the CoM method. We address each major comment below and indicate the revisions made to the manuscript.

read point-by-point responses

  1. Referee: [Results section] Results section (examples of increasing complexity): the manuscript asserts robustness and accuracy but provides no quantitative propagation analysis showing how local committor errors—especially near the dividing surface—translate into MFPT bias in the Milestoning flux integrals. Short trajectories may undersample transition-state regions in high-barrier cases, and this risk is not explicitly bounded.

    Authors: We agree that an explicit quantitative propagation analysis would strengthen the presentation. The original manuscript validated overall MFPT accuracy via direct comparison to reference long-trajectory results across the test systems of increasing complexity. In the revised version we have added a dedicated subsection to the Results that performs a sensitivity analysis: we introduce controlled Gaussian perturbations to the learned committor specifically in the vicinity of the dividing surface, recompute the Milestoning flux integrals, and report the resulting MFPT bias together with empirical error bounds derived from the observed committor RMSE. This analysis shows that the induced MFPT error remains below 10 % for the barriers examined; we also note the increased sampling requirement for extremely high barriers and have added a brief discussion of this limitation. revision: yes

  2. Referee: [Methods section] Methods section (analogue prediction step): the procedure for extending committor values from short trajectories is described at a high level, yet no error metrics, bias controls, or sensitivity tests are reported that would confirm the neural-network surface remains sufficiently accurate for subsequent Milestoning.

    Authors: We acknowledge that the analogue-prediction procedure was presented at a high level in the original submission. In the revised Methods section we now report quantitative error metrics for the neural-network committor: root-mean-square error on a held-out validation set of short trajectories, five-fold cross-validation scores, and sensitivity of the final MFPT to the number and length of the short trajectories used for training. These controls confirm that the committor surface is accurate to within approximately 0.05 on average, which is sufficient to keep the Milestoning flux integrals within the tolerance demonstrated by the reference comparisons. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation relies on independent simulations and standard Milestoning

full rationale

The paper's chain proceeds from short parallelizable trajectories plus analogue prediction to train an NN representation of the committor, which is then fed into the established Milestoning procedure to obtain MFPT. None of these steps reduce the final MFPT value to a fitted parameter or self-referential definition by construction; the short-trajectory data and Milestoning fluxes remain external to the target observable. No self-citation load-bearing, uniqueness theorems, or ansatz smuggling appear in the provided description, and the central claim is an empirical assertion about computational efficiency and accuracy rather than a tautological re-expression of inputs.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Ledger is provisional because only the abstract is available; free parameters are the neural-network weights and any analogue-prediction hyperparameters; axioms include the assumption that local short trajectories suffice for global committor accuracy.

free parameters (1)
  • neural network parameters
    Weights and biases of the neural network ansatz for the committor are fitted to short-trajectory data.
axioms (1)
  • domain assumption Short trajectory simulations capture sufficient local kinetic information to enable accurate global committor estimation via analogue prediction.
    Invoked in the description of the fast estimation step.

pith-pipeline@v0.9.0 · 5653 in / 1246 out tokens · 30779 ms · 2026-05-20T01:25:21.980889+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

53 extracted references · 53 canonical work pages

  1. [1]

    Science334(6055), 517–520 (2011) https://doi.org/10.1126/science.1208351

    Lindorff-Larsen, K., Piana, S., Dror, R.O., Shaw, D.E.: How fast-folding proteins fold. Science334(6055), 517–520 (2011) https://doi.org/10.1126/science.1208351

    work page doi:10.1126/science.1208351 2011

  2. [2]

    Liu, Y., Wang, R., Wang, H., Liu, W.: Unbinding dynamics of large hiv non- nucleoside reverse transcriptase inhibitors revealed by atomistic milestoning simulations. J. Phys. Chem. B129(47), 12076–12086 (2025) https://doi.org/10. 1021/acs.jpcb.5c04843

    work page 2025

  3. [3]

    Living J

    H´ enin, J., L` elievre, T., Shirts, M.R., Valsson, O., Delemotte, L.: Enhanced sam- pling methods for molecular dynamics simulations. Living J. Comp. Mol. Sci. 4(1), 1583 (2022) https://doi.org/10.33011/livecoms.4.1.1583

    work page doi:10.33011/livecoms.4.1.1583 2022

  4. [4]

    WIREs Comput

    Elber, R., Fathizadeh, A., Ma, P., Wang, H.: Modeling molecular kinetics with milestoning. WIREs Comput. Mol. Sci.11(4), 1512 (2021) https://doi.org/10. 1002/wcms.1512 18

    work page 2021

  5. [5]

    Drug Discov

    Pan, A.C., Borhani, D.W., Dror, R.O., Shaw, D.E.: Molecular determinants of drug-receptor binding kinetics. Drug Discov. Today18(13), 667–673 (2013) https: //doi.org/10.1016/j.drudis.2013.02.007

    work page doi:10.1016/j.drudis.2013.02.007 2013

  6. [6]

    Copeland, R.A.: The drug-target residence time model: A 10-year retrospective. Nat. Rev. Drug Discov.15(2), 87–95 (2016) https://doi.org/10.1038/nrd.2015.18

    work page doi:10.1038/nrd.2015.18 2016

  7. [8]

    Husic, B.E., Pande, V.S.: Markov state models: From an art to a science. J. Am. Chem. Soc.140(7), 2386–2396 (2018) https://doi.org/10.1021/jacs.7b12191

    work page doi:10.1021/jacs.7b12191 2018

  8. [9]

    van Erp, T.S., Bolhuis, P.G.: Elaborating transition interface sampling methods. J. Comput. Phys.205(1), 157–181 (2005) https://doi.org/10.1016/j.jcp.2004.11. 003

    work page doi:10.1016/j.jcp.2004.11 2005

  9. [10]

    Allen, R.J., Frenkel, D., Wolde, P.R.: Forward flux sampling-type schemes for simulating rare events: Efficiency analysis. J. Chem. Phys.124(19), 194111 (2006) https://doi.org/10.1063/1.2198827

    work page doi:10.1063/1.2198827 2006

  10. [11]

    Allen, R.J., Valeriani, C., Wolde, P.R.: Forward flux sampling for rare event sim- ulations. J. Phys.: Condens. Matter21(46), 463102 (2009) https://doi.org/10. 1088/0953-8984/21/46/463102

    work page 2009

  11. [12]

    Zhang, B.W., Jasnow, D., Zuckerman, D.M.: The weighted ensemble path sam- pling method is statistically exact for a broad class of stochastic processes and binning procedures. J. Chem. Phys.132(5), 054107 (2010) https://doi.org/10. 1063/1.3306345

    work page 2010

  12. [13]

    C´ erou, F., Guyader, A.: Adaptive multilevel splitting for rare event anal- ysis. Stoch. Anal. Appl.25(2), 417–443 (2007) https://doi.org/10.1080/ 07362990601139628

    work page 2007

  13. [14]

    Lopes, L.J.S., Leli` evre, T.: Analysis of the adaptive multilevelsplitting method on the iosmerizatio of alanine dipeptide. J. Comput. Chem.40(11), 1198–1208 (2019) https://doi.org/10.1002/jcc.25778

    work page doi:10.1002/jcc.25778 2019

  14. [15]

    Faradjian, A.K., Elber, R.: Computing time scales from reaction coordinates by milestoning. J. Chem. Phys.120(23), 10880–10889 (2004) https://doi.org/10. 1063/1.1738640 19

    work page 2004

  15. [16]

    Bello-Rivas, J.M., Elber, R.: Exact milestoning. J. Chem. Phys.142(9), 094102 (2015) https://doi.org/10.1063/1.4913399

    work page doi:10.1063/1.4913399 2015

  16. [17]

    E, W., Vanden-Eijnden, E.: Towards a theory of transition paths. J. Stat. Phys. 123(3), 503–523 (2006) https://doi.org/10.1007/s10955-005-9003-9

    work page doi:10.1007/s10955-005-9003-9 2006

  17. [18]

    Multiscale Model

    Metzner, P., Sch¨ utte, C., Vanden-Eijnden, E.: Transition path theory for markov jump processes. Multiscale Model. Simul.7(3), 1192–1219 (2009) https://doi.org/ 10.1137/070699500

    work page doi:10.1137/070699500 2009

  18. [19]

    Krivov, S.V.: On reaction coordinate optimality. J. Chem. Theory Comput.9(1), 135–146 (2013) https://doi.org/10.1021/ct3008292

    work page doi:10.1021/ct3008292 2013

  19. [20]

    Vanden-Eijnden, E., Venturoli, M., Ciccotti, G., Elber, R.: On the assumptions underlying milestoning. J. Chem. Phys.129(17), 174102 (2008) https://doi.org/ 10.1063/1.2996509

    work page doi:10.1063/1.2996509 2008

  20. [21]

    Lucente, D., Rolland, J., Herbert, C., Bouchet, F.: Coupling rare event algorithms with data-based learned committor functions using the analogue markov chain. J. Stat. Mech.2022(8), 083201 (2022) https://doi.org/10.1088/1742-5468/ac7aa7

    work page doi:10.1088/1742-5468/ac7aa7 2022

  21. [22]

    CRC Press, Boca Raton (1996)

    Durrett, R.: Stochastic Calculus: A Practical Introduction. CRC Press, Boca Raton (1996)

    work page 1996

  22. [23]

    Khoo, Y., Lu, J., Ying, L.: Solving for high-dimensional committor functions using artificial neural networks. Res. Math. Sci.6(1), 1 (2018) https://doi.org/10.1007/ s40687-018-0160-2

    work page 2018

  23. [24]

    Li, Q., Lin, B., Ren, W.: Computing committor functions for the study of rare events using deep learning. J. Chem. Phys.151(5), 054112 (2019) https://doi. org/10.1063/1.5110439

    work page doi:10.1063/1.5110439 2019

  24. [25]

    Journal of Computational Physics544, 114439 (2026) https://doi.org/10.1016/j.jcp.2025

    Wang, Y., Tang, K., Wang, X., Wan, X., Ren, W., Yang, C.: Estimating com- mittor functions via deep adaptive sampling on rare transition paths. Journal of Computational Physics544, 114439 (2026) https://doi.org/10.1016/j.jcp.2025. 114439

    work page doi:10.1016/j.jcp.2025 2026

  25. [26]

    Kang, P., Trizio, E., Parrinello, M.: Computing the committor with the committor to study the transition state ensemble. Nat. Comput. Sci.4(6), 451–460 (2024) https://doi.org/10.1038/s43588-024-00645-0

    work page doi:10.1038/s43588-024-00645-0 2024

  26. [27]

    Banushkina, P.V., Krivov, S.V.: Nonparametric variational optimization of reac- tion coordinates. J. Chem. Phys.143(18), 184108 (2015) https://doi.org/10.1063/ 1.4935180 20

    work page 2015

  27. [28]

    Strahan, J., Finkel, J., Dinner, A.R., Weare, J.: Predicting rare events using neural networks and short-trajectory data. J. Comput. Phys.488, 112152 (2023) https: //doi.org/10.1016/j.jcp.2023.112152

    work page doi:10.1016/j.jcp.2023.112152 2023

  28. [29]

    He, Z., Chipot, C., Roux, B.: Committor-consistent variational string method. J. Phys. Chem. Lett.13(40), 9263–9271 (2022) https://doi.org/10.1021/acs.jpclett. 2c02529

    work page doi:10.1021/acs.jpclett 2022

  29. [30]

    Chen, H., Roux, B., Chipot, C.: Discovering reaction pathways, slow variables, and committor probabilities with machine learning. J. Chem. Theory Comput. 19(14), 4414–4426 (2023) https://doi.org/10.1021/acs.jctc.3c00028

    work page doi:10.1021/acs.jctc.3c00028 2023

  30. [31]

    Ji, X., Wang, R., Wang, H., Liu, W.: On committor functions in milestoning. J. Chem. Phys.159(24), 244115 (2023) https://doi.org/10.1063/5.0180513

    work page doi:10.1063/5.0180513 2023

  31. [32]

    Lguensat, R., Tandeo, P., Ailliot, P., Pulido, M., Fablet, R.: The analog data assimilation. Mon. Weather Rev.145(10), 4093–4107 (2017) https://doi.org/10. 1175/MWR-D-16-0441.1

    work page 2017

  32. [33]

    JAS78(7), 2117–2133 (2021) https://doi.org/10.1175/JAS-D-20-0204.1

    Platzer, P., Yiou, P., Naveau, P., Tandeo, P., Filipot, J.-F., Ailliot, P., Zhen, Y.: Using local dynamics to explain analog forecasting of chaotic systems. JAS78(7), 2117–2133 (2021) https://doi.org/10.1175/JAS-D-20-0204.1

    work page doi:10.1175/jas-d-20-0204.1 2021

  33. [34]

    Non- linear Processes in Geophysics30(2), 195–216 (2023) https://doi.org/10.5194/ npg-30-195-2023

    Jacques-Dumas, V., Westen, R.M., Bouchet, F., Dijkstra, H.A.: Data-driven methods to estimate the committor function in conceptual ocean models. Non- linear Processes in Geophysics30(2), 195–216 (2023) https://doi.org/10.5194/ npg-30-195-2023

    work page 2023

  34. [35]

    M´ ajek, P., Elber, R.: Milestoning without a reaction coordinate. J. Chem. Theory Comput.6(6), 1805–1817 (2010) https://doi.org/10.1021/ct100114j

    work page doi:10.1021/ct100114j 2010

  35. [36]

    Wang, R., Wang, H., Liu, W., Elber, R.: Approximating first hitting point distri- bution in milestoning for rare event kinetics. J. Chem. Theory Comput.19(19), 6816–6826 (2023) https://doi.org/10.1021/acs.jctc.3c00315

    work page doi:10.1021/acs.jctc.3c00315 2023

  36. [37]

    Narayan, B., Elber, R.: Comparison of accuracy and efficiency of milestoning variants: Introducing buffer milestoning. J. Phys. Chem. B128(6), 1438–1447 (2024) https://doi.org/10.1021/acs.jpcb.3c07933

    work page doi:10.1021/acs.jpcb.3c07933 2024

  37. [38]

    BAMS50(5), 345– 349 (1969)

    Lorenz, E.N.: Three approaches to atmospheric predictability. BAMS50(5), 345– 349 (1969)

    work page 1969

  38. [39]

    JAS26(4), 636–646 (1969) https://doi.org/10.1175/1520-0469(1969) 26⟨636:APARBN⟩2.0.CO;2

    Lorenz, E.N.: Atmospheric predictability as revealed by naturally occurring 21 analogues. JAS26(4), 636–646 (1969) https://doi.org/10.1175/1520-0469(1969) 26⟨636:APARBN⟩2.0.CO;2

    work page doi:10.1175/1520-0469(1969 1969

  39. [40]

    Laio, A., Parrinello, M.: Escaping free-energy minima. Proc. Natl. Acad. Sci. U.S.A.99(20), 12562–12566 (2002) https://doi.org/10.1073/pnas.202427399

    work page doi:10.1073/pnas.202427399 2002

  40. [41]

    Barducci, A., Bussi, G., Parrinello, M.: Well-tempered metadynamics: A smoothly converging and tunable free-energy method. Phys. Rev. Lett.100, 020603 (2008) https://doi.org/10.1103/PhysRevLett.100.020603

    work page doi:10.1103/physrevlett.100.020603 2008

  41. [42]

    Sugita, Y., Okamoto, Y.: Replica-exchange molecular dynamics method for pro- tein folding. Chem. Phys. Lett.314(1), 141–151 (1999) https://doi.org/10.1016/ S0009-2614(99)01123-9

    work page 1999

  42. [43]

    Leimkuhler, B., Matthews, C.: Rational construction of stochastic numerical methods for molecular sampling. Appl. Math. Res. eXpress2013(1), 34–56 (2013) https://doi.org/10.1093/amrx/abs010

    work page doi:10.1093/amrx/abs010 2013

  43. [44]

    Mitchell, A.R., Rotskoff, G.M.: Committor guided estimates of molecular transi- tion rates. J. Chem. Theory Comput.20(21), 9378–9393 (2024) https://doi.org/ 10.1021/acs.jctc.4c00997

    work page doi:10.1021/acs.jctc.4c00997 2024

  44. [45]

    Rogal, J.: Reaction coordinates in complex systems-a perspective. Eur. Phys. J. B94(11), 223 (2021) https://doi.org/10.1140/epjb/s10051-021-00233-5

    work page doi:10.1140/epjb/s10051-021-00233-5 2021

  45. [46]

    Phillips, J.C., Hardy, D.J., Maia, J.D.C., Stone, J.E., Ribeiro, J.a.V., Bernardi, R.C., Buch, R., Fiorin, G., H´ enin, J., Jiang, W., McGreevy, R., Melo, M.C.R., Radak, B.K., Skeel, R.D., Singharoy, A., Wang, Y., Roux, B., Aksimentiev, A., Luthey-Schulten, Z., Kal´ e, L.V., Schulten, K., Chipot, C., Tajkhorshid, E.: Scal- able molecular dynamics on cpu a...

    work page doi:10.1063/5.0014475 2020

  46. [47]

    MacKerell, A.D.J., Bashford, D., Bellott, M., Dunbrack, R.L.J., Evanseck, J.D., Field, M.J., Fischer, S., Gao, J., Guo, H., Ha, S., Joseph-McCarthy, D., Kuchnir, L., Kuczera, K., Lau, F.T.K., Mattos, C., Michnick, S., Ngo, T., Nguyen, D.T., Prodhom, B., Reiher, W.E., Roux, B., Schlenkrich, M., Smith, J.C., Stote, R., Straub, J., Watanabe, M., Wi´ orkiewic...

    work page doi:10.1021/jp973084f 1998

  47. [48]

    Honda, S., Akiba, T., Kato, Y.S., Sawada, Y., Sekijima, M., Ishimura, M., Ooishi, A., Watanabe, H., Odahara, T., Harata, K.: Crystal structure of a ten-amino acid protein. J. Am. Chem. Soc.130(46), 15327–15331 (2008) https://doi.org/ 22 10.1021/ja8030533

    work page doi:10.1021/ja8030533 2008

  48. [49]

    Martyna, G.J., Tobias, D.J., Klein, M.L.: Constant pressure molecular dynamics algorithms. J. Chem. Phys.101(5), 4177–4189 (1994) https://doi.org/10.1063/1. 467468

    work page doi:10.1063/1 1994

  49. [50]

    Feller, S.E., Zhang, Y., Pastor, R.W., Brooks, B.R.: Constant pressure molecu- lar dynamics simulation: The Langevin piston method. J. Chem. Phys.103(11), 4613–4621 (1995) https://doi.org/10.1063/1.470648

    work page doi:10.1063/1.470648 1995

  50. [51]

    Miyamoto, S., Kollman, P.A.: Settle: An analytical version of the shake and rattle algorithm for rigid water models. J. Comput. Chem.13(8), 952–962 (1992) https: //doi.org/10.1002/jcc.540130805

    work page doi:10.1002/jcc.540130805 1992

  51. [52]

    Ryckaert, J.-P., Ciccotti, G., Berendsen, H.J.C.: Numerical integration of the cartesian equations of motion of a system with constraints: molecular dynamics of n-alkanes. J. Comput. Phys.23(3), 327–341 (1977) https://doi.org/10.1016/ 0021-9991(77)90098-5

    work page 1977

  52. [53]

    Essmann, U., Perera, L., Berkowitz, M.L., Darden, T., Lee, H., Pedersen, L.G.: A smooth particle mesh Ewald method. J. Chem. Phys.103(19), 8577–8593 (1995) https://doi.org/10.1063/1.470117

    work page doi:10.1063/1.470117 1995

  53. [54]

    Best, R.B., Zhu, X., Shim, J., Lopes, P.E.M., Mittal, J., Feig, M., MacKerell, A.D.J.: Optimization of the additive charmm all-atom protein force field target- ing improved sampling of the backboneϕ,ψand side-chainχ1 andχ2 dihedral angles. J. Chem. Theory Comput.8(9), 3257–3273 (2012) https://doi.org/10. 1021/ct300400x 6 Acknowledgements This work was sup...

    work page 2012