- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
The paper learns a large-timestep integrator for Langevin dynamics. Given the current state and the random noise path over the step, a network predicts the next position and velocity for steps of up to about 10 fs. The noise enters exactly, and the network learns only the time-averaged velocity and force. Training uses shuffled equilibrium configurations and forces rather than time-ordered trajectories. It combines force and velocity regression at zero step size with a self-consistency loss that ties one step of size h to two steps of size h/2. This extends recent stochastic flow-map work (McCallum et al.) to Langevin dynamics. Tests cover small molecules (MD17), the kinetics of alanine dipeptide, and a transferable model for short peptides.
Critical strengths
The design is clean. Learning the Langevin step directly is meant to remove the splitting bias that Hamiltonian flow maps have when combined with a separate noise step. The paper argues this but does not test it head-to-head on dynamics. On small molecules the accuracy is convincing. On ethanol, a 10 fs step is about as accurate as a standard integrator at about 1 fs (Fig. 2). Structural errors are below both the earlier Hamiltonian flow-map method and the reference ML force field (Fig. 3). In the peptide setting, large steps beat Amber14 on state-occupancy and transition-probability error in 10 of 12 cases at 4 and 8 fs (Table 3).
Critical weaknesses
- The practical speed advantage isn't shown. The abstract's "order of magnitude faster than current machine-learned force fields" is real, but Fig. 5 shows it is simply the 10 fs vs 1 fs step ratio (App. C.2). Against classical MD on the same molecule it is only about 13% faster. For larger peptides, LFM at 8 fs becomes slower than classical MD at 1 fs beyond about 70 atoms (Fig. 14). No baseline uses the 2–4 fs steps common in practice, and training wall-clock time isn't reported. The convergence comparisons (Figs. 4 and 12, Table 3) are made at equal network evaluations, which is not equal wall-clock, since one LFM step costs several classical steps (Fig. 14).
- The large-step kinetics (transition times and fluxes) are checked on one small system, and the peptide model is biased. Transition times and fluxes are compared with MD only for alanine dipeptide, with no error bars, and all three LFM transition times come out about 6–8% shorter than MD (Fig. 5a). For peptides, the only dynamics metrics are averaged over three peptides per length with no spread (Table 3). At 8 fs the energy distribution is 6–12× further from the reference than standard MD (torsions 1.2–2.3×). There is no Metropolis or reweighting step to remove this bias. An inference-time temperature corrector is used (App. C.3.1), but the main text never mentions it. The authors propose exploring with large steps and then relaxing with small ones (Sec. 4.2), which is reasonable, but it is shown only as a single-peptide example (Fig. 13), with no comparison of discovery rates against plain MD at matched cost. Sec. 4.2's "consistently outperform" also fails in 2 of the 12 large-step cells.
Overall: Convincing single-molecule accuracy, but the practical speed advantage and large-step accuracy on peptides aren't yet established. Error bars, plus claims narrowed to what is shown, would move me to 3; so would a wall-clock win over tuned classical MD.
Questions for authors
- Can you compare wall-clock time against classical MD at 2–4 fs, and report training time?
- Can you add error bars to Fig. 5 and Table 3? Are the shorter transition times outside MD sampling error, and what friction does the reference MD use?
- Table 3 compares at equal network evaluations (10⁷), so the 8 fs model simulates 8× more time. How do the results compare at equal simulated time and at equal wall-clock?
- Which results use the temperature corrector? App. C.3.1 says the main correction just scales the noise, with a fallback to velocity rescaling, but gives no α values, no fallback rate, and no effect on transition statistics.
- Fig. 13: the new state survives 7–16 ns in three Amber14 runs but is absent from the reference. Can you estimate its equilibrium population, to rule out an artefact of the biased large-step ensemble?
3 comments
Response to xtuy - part 2
Table R9: Iso-simulation time transferable LFM sampling.
| Forces | Integrator | Timestep | 2AA E-W₂ | 2AA T-W₂ | 2AA π-TV | 2AA P-TV | 4AA E-W₂ | 4AA T-W₂ | 4AA π-TV | 4AA P-TV | 8AA E-W₂ | 8AA T-W₂ | 8AA π-TV | 8AA P-TV |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LFM | LFM | 8 fs | 3.58 | 0.84 | 0.13 | 0.15 | 11.93 | 2.33 | 0.25 | 0.36 | 25.92 | 4.19 | 0.06 | 0.54 |
| LFM | LFM | 4 fs | 2.76 | 0.64 | 0.13 | 0.15 | 8.58 | 2.09 | 0.19 | 0.31 | 21.31 | 4.29 | 0.10 | 0.52 |
| LFM | LFM | 2 fs | 1.86 | 0.61 | 0.24 | 0.14 | 7.13 | 1.95 | 0.21 | 0.30 | 17.78 | 4.16 | 0.17 | 0.48 |
| LFM | LFM | 1 fs | 1.02 | 0.63 | 0.30 | 0.19 | 5.74 | 1.98 | 0.21 | 0.33 | 15.33 | 4.37 | 0.16 | 0.45 |
| LFM | BAOAB | 1 fs | 0.31 | 0.37 | 0.19 | 0.11 | 0.97 | 1.62 | 0.41 | 0.31 | 3.04 | 2.90 | 0.07 | 0.48 |
| Amber14 | BAOAB | 1 fs | 0.26 | 0.33 | 0.15 | 0.12 | 1.20 | 1.26 | 0.24 | 0.33 | 2.38 | 3.13 | 0.13 | 0.47 |
Table R10: Iso-wallclock time transferable LFM sampling.
| Forces | Integrator | Timestep | 2AA E-W₂ | 2AA T-W₂ | 2AA π-TV | 2AA P-TV | 4AA E-W₂ | 4AA T-W₂ | 4AA π-TV | 4AA P-TV | 8AA E-W₂ | 8AA T-W₂ | 8AA π-TV | 8AA P-TV |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LFM | LFM | 8 fs | 3.58 | 0.81 | 0.10 | 0.11 | 11.93 | 2.31 | 0.24 | 0.26 | 25.92 | 4.10 | 0.04 | 0.43 |
| LFM | LFM | 4 fs | 2.77 | 0.63 | 0.08 | 0.12 | 8.57 | 2.05 | 0.11 | 0.23 | 21.31 | 4.19 | 0.07 | 0.38 |
| LFM | LFM | 2 fs | 1.85 | 0.60 | 0.23 | 0.07 | 7.23 | 1.94 | 0.12 | 0.22 | 18.10 | 4.05 | 0.16 | 0.42 |
| LFM | LFM | 1 fs | 1.02 | 0.64 | 0.29 | 0.11 | 5.90 | 1.95 | 0.17 | 0.31 | 15.12 | 4.33 | 0.15 | 0.46 |
| LFM | BAOAB | 1 fs | 0.35 | 0.34 | 0.13 | 0.14 | 0.92 | 1.45 | 0.35 | 0.28 | 2.77 | 2.84 | 0.07 | 0.48 |
| Amber14 | BAOAB | 1 fs | 0.14 | 0.20 | 0.06 | 0.04 | 0.60 | 0.91 | 0.15 | 0.19 | 1.21 | 2.34 | 0.12 | 0.46 |
- Which results use the temperature corrector? App. C.3.1 says the main correction just scales the noise, with a fallback to velocity rescaling, but gives no α values, no fallback rate, and no effect on transition statistics.
Only the transferable LFM model uses the temperature corrector during inference. We have measured the values and fallback rate which are shown in Table 7. Reproduced below in Table R11 for convenience. The median value reduces the standard deviation of the noise and falls back to velocity rescaling around one third of the time.
Table R11: Temperature corrector statistics.
| Timestep | α median [IQR] | Fallback |
|---|---|---|
| 1 fs | 0.45 [−0.32, 1.57] | 33% |
| 2 fs | 0.48 [−0.18, 1.33] | 34% |
| 4 fs | 0.43 [−0.10, 1.06] | 34% |
| 8 fs | 0.34 [−0.10, 0.80] | 37% |
- Fig. 13: the new state survives 7–16 ns in three Amber14 runs but is absent from the reference. Can you estimate its equilibrium population, to rule out an artefact of the biased large-step ensemble?
We have estimated the meta-stable state's equilibrium population using the forward and reverse estimation method from Klimovich et al. (2015) using replica-exchange MD. The result is shown in Figure 17 and predicts an occupancy of %. This shows that the state is rare but stable and present in the classical MD Boltzmann distribution.
Concluding remarks Overall, we believe that the contribution of LFMs is to improve the runtime of MLFFs while maintaining comparable accuracy. We agree that the runtime improvement over tuned classical-force-field MD is not yet established and have updated the manuscript to further reflect this in the integration throughput discussion, conclusions and limitations and future work sections.
We thank the reviewer again for their feedback which has strengthened the contributions of the paper. If there are any further questions, we are happy to continue the discussion.
Dear reviewer xtuy,
We are thankful for your time and thoughtful review, especially the suggestion for further comparison to classical MD force-fields which we have now added.
As the end of the rebuttal period approaches, we are happy to answer any further questions. If our responses and additional experiments have addressed your questions, we would be grateful if you might consider updating your assessment.
Response to xtuy - part 1
Thank you for your review and suggestions for improvement. We reply to each question in turn below.
We have measured the integration throughput of classical MD at 2-4fs, using hydrogen bond constraints at 4fs. Please see Figure 18, or Table R6 below for a summary. The throughput of classical MD is higher than the LFM in the 2-4fs timestep range. The LFM obtains a slightly higher throughput at 8fs compared to OpenMM at 1fs for small and medium systems.
The transferable LFM does not obtain a speedup over classical MD. This is now highlighted within the text in the integration throughput discussion, conclusions and future work sections. We note that the LFMs do obtain a significant speedup over MLFFs and this is the application where we hope LFMs can provide a contribution.
To further validate this improvement over MLFFs and previous Hamiltonian Flow Map (HFM) architectures, we have trained the LFM and HFM baseline up to 50fs steps on alanine dipeptide. We report the transition dynamics predicted by the HFM and LFM models in Figure 13. At 50fs steps, the LFM obtains accurate transition probabilities between meta-stable states within +/- 2pp, while the HFM has diverged due to integration bias. This suggests the possibility for scaling the LFM model to much larger timesteps to obtain further integration speedups over MLFFs.
Table R6: Integration throughput in ns/hour.
The training time for the transferable LFM is reported in Table R7 below.
Table R7: Transferable training time.
Error bars have been added to Figure 5 and Table 3. The reference MD uses the same friction of γ=10ps$^{-1}$ as the LFM. See Table R8 for the uncertainty estimates for the alanine dipeptide MFPTs. The shorter transition times of the LFM are slightly outside MD sampling error.
Table R8: Alanine dipeptide MFPTs plus 95% uncertainty ranges.
We have computed the transferable LFM Table 3 results for iso-simulation and iso-wallclock time. Please see Tables 8 and 9, reproduced below in Tables R9 and R10 respectively (comment 2).
At iso-simulation time, the large-timestep LFM maintains the improved meta-stable state convergence across all system sizes and comparable transition dynamics accuracy for 4AA and 8AA systems.
At iso-wallclock time, the large-timestep LFM achieves improved meta-stable state convergence across the 4AA and 8AA systems and is comparable to the Amber14 force-field on the 2AA systems.
We believe that these are already promising results as a demonstration of a transferable MLFF that is capable of large-timestep integration. However, we agree that more work is required to obtain an integrator with comparable performance to classical MD. This sentiment is reflected in the integration throughput discussion, conclusions and limitations and future work section.