- Rating
- 3(+1)
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
The authors propose a second-order neural ordinary differential equation framework for temporal scRNA-seq data. Previous methods rely only on the current state and time to learn first-order dynamics. The authors argue that those two are not sufficient, motivated by three aspects: (i) the equations of classical splicing models result in second-order dynamics with respect to the current state, (ii) when high-dimensional data are projected into a low-dimensional setup, the order of dynamics should increase to provide sufficient information about the gene space, and finally (iii) similar cells measured at a time do not mean that their future will be similar. To address these limitations, the authors augment the dynamics model by adding velocity information that allows the model to learn history-dependent dynamics. They estimate the initial velocity from the first two snapshots by utilizing optimal transport.
Critical strengths
- The paper raises an important point: most existing approaches infer dynamics from the current position alone. The authors motivate, from biophysical, mathematical, and empirical perspectives, why information about history beyond position may be needed.
- The authors evaluate on diverse datasets against established baselines. The method achieves the best results in most settings (Table 1).
- Strong experimental design supports the paper’s claims. Section 4.3 tests Proposition 1 by projecting first-order gene-space dynamics onto fewer dimensions, showing that first-order closure breaks under compression while second-order largely restores it at the model’s latent dimension. Section 4.4 tests Proposition 2 on synthetic data with known coupling, where the first-order reduction fails as the theory predicts. The ablation in Section 4.6 separates the contributions of the velocity state and the OT initialization.
Critical weaknesses
- Some related works are not discussed. While they do not address history-dependent dynamics, they target similar interpolation and extrapolation tasks and could be strong competitors:
- [1] von Bassewitz, J. P., Pervez, A., Fumero, M., Robinson, M., Karaletsos, T., & Locatello, F. (2026). Learning explicit single-cell dynamics using ODE representations. In International Conference on Learning Representations.
- [2] Balik, M. Y., & Lähdesmäki, H. (2026). Modeling Temporal scRNA-seq Data with Latent Gaussian Process and Optimal Transport. In Forty-Third International Conference on Machine Learning.
- Proposition 3 requires to be uniformly Lipschitz in with constant and . Yet, nothing in training enforces to be larger than . Is this condition always guaranteed to hold? It is not clear that the proposition applies to the trained model.
- The paper does not report the learned or for the benchmark datasets. Reporting and per benchmark dataset and analyzing it to show whether velocity has time to depart from the value set by position and time would strengthen the paper.
Questions for authors
-
I have several concerns regarding the baselines: In the original papers, some baselines were designed specifically for PCA space, and some can handle 2000-dimensional gene space. Could the authors please clarify how they trained the baselines (i.e., on PCA space or gene space) in this work? Also, were any hyperparameters tuned for the baselines and their model? If so, please specify the search space and validation setup.
-
Could the authors please elaborate on why the first-order ablation variant performs substantially better than scNODE on SC extrapolation (53 vs 88 ) despite the general idea and architecture being similar?
-
How does the runtime (both training and inference) compare against the baselines?
-
The temporal prediction experiment considers interpolation-only and extrapolation-only setups. It would be nice to see high-sparsity (hard tasks in the scNODE paper) experiments that contain evaluation on both interpolation and extrapolation time points together.
5 comments
Weakness 2
Proposition 3 requires f_\theta to be uniformly Lipschitz in v with constant \Lambda and \gamma>\Lambda. Yet, nothing in training enforces \gamma to be larger than \Lambda. Is this condition always guaranteed to hold? It is not clear that the proposition applies to the trained model.
The paper does not report the learned \gamma or \tau = frac{1}{\gamma} for the benchmark datasets. Reporting \gamma and \Lambda per benchmark dataset and analyzing it to show whether velocity has time to depart from the value set by position and time would strengthen the paper.
Is the condition guaranteed? No. Nothing in training enforces , and we leave it unenforced on purpose. Proposition 3 is a statement about the model class: it shows that the second-order model contains first-order dynamics as its overdamped regime. Enforcing the condition would push scMomenta toward this regime, and with a large margin it would reduce the model to a first-order one, removing the history dependence that motivates the paper. Since neither training nor prediction requires the condition, we learn freely and let the data decide how far the model departs from the first-order limit.
Learned and . Following the reviewer's suggestion, Table R3-2 reports , and of the trained models. is the time scale over which damping erases the velocity, that is, how long the history carried by the velocity persists; we report it in units of the median interval between consecutive observed time points. Since the global Lipschitz constant cannot be computed, the local is the median, over points on the simulated trajectories, of the spectral norm of , and we also report the fraction of these points at which . We have added this table to Appendix H.4.
Table R3-2. Learned damping of scMomenta; mean ± s.d. over five seeds.
| Dataset | Task | Local (median) | Points with | ||
|---|---|---|---|---|---|
| ZB | Interp. | 0.36 ± 0.02 | 2.80 | 4.70 ± 1.10 | 0% |
| ZB | Extrap. | 0.24 ± 0.01 | 4.23 | 3.62 ± 0.85 | 0% |
| SC | Interp. | 0.45 ± 0.03 | 2.21 | 6.21 ± 0.19 | 0% |
| SC | Extrap. | 0.50 ± 0.08 | 2.03 | 9.17 ± 0.64 | 0% |
| DR | Interp. | 5.16 ± 0.09 | 0.19 | 9.74 ± 0.74 | 8% |
| DR | Extrap. | 4.89 ± 0.12 | 0.20 | 12.49 ± 2.61 | 4% |
| GA | Interp. | 1.82 ± 0.16 | 0.55 | 3.02 ± 2.11 | 33% |
| GA | Extrap. | 1.60 ± 0.01 | 0.63 | 1.15 ± 0.95 | 70% |
| MA | Interp. | 0.82 ± 0.30 | 1.32 | 3.97 ± 3.50 | 0% |
| MA | Extrap. | 0.96 ± 0.27 | 1.12 | 6.74 ± 3.92 | 0% |
The learned damping varies widely across datasets, from on ZB to 5.16 on DR. On ZB, SC and MA, exceeds one observation interval, so the velocity is not set by the current position alone, and the dynamics depend on each cell's history. On DR, is only a fifth of an interval and the model is nearly first order; GA lies in between. This agrees with the temporal prediction results (Table 2): on DR, a first-order method (scIMF) achieves a lower extrapolation than scMomenta, as does scNODE on GA.
Question 1
I have several concerns regarding the baselines: In the original papers, some baselines were designed specifically for PCA space, and some can handle 2000-dimensional gene space. Could the authors please clarify how they trained the baselines (i.e., on PCA space or gene space) in this work? Also, were any hyperparameters tuned for the baselines and their model? If so, please specify the search space and validation setup.
Representation space. Each baseline was trained in the representation used in its original paper and official implementation, for example scNODE in the latent space of its VAE, and PI-SDE and scIMF in a PCA space. Every training representation, whether a PCA or an autoencoder, was fitted on the cells of the observed time points only, and the predictions of all methods were mapped back to gene space and evaluated with the same pipeline.
Hyperparameters. Each baseline was run from its official implementation with the settings recommended by its authors, without additional tuning. scMomenta uses a single setting for all five datasets, selected using the observed time points only. Held-out time points were never used for model selection or for fitting any training representation.
Question 2
Could the authors please elaborate on why the first-order ablation variant performs substantially better than scNODE on SC extrapolation (53 vs 88 W_2) despite the general idea and architecture being similar?
Where the gap arises. The two models differ substantially only on SC extrapolation (Table R3-3a).
Table R3-3a. of scNODE (Tables 1 and 2) and the first-order ablation (Table 3); mean ± s.d. over five seeds.
| Method | ZB interp. | ZB extrap. | SC interp. | SC extrap. |
|---|---|---|---|---|
| scNODE | 28.89 ± 1.08 | 48.68 ± 1.04 | 33.33 ± 0.98 | 88.75 ± 22.5 |
| First-order ablation | 28.37 ± 0.28 | 42.55 ± 1.01 | 31.95 ± 0.30 | 53.32 ± 2.07 |
This is also the only setting in which scNODE is unstable across seeds (Table R3-3b): its coefficient of variation there is 25%, versus at most 9.2% in its other nine settings of Tables 1 and 2, whereas that of the first-order ablation is 3.9%. The gap is therefore specific to scNODE on this task, not to first-order latent dynamics or to our evaluation pipeline.
Table R3-3b. of scNODE on SC extrapolation for each seed.
| Seed | 42 | 43 | 44 | 45 | 46 | Mean ± s.d. |
|---|---|---|---|---|---|---|
| 84.66 | 66.49 | 121.13 | 100.63 | 70.82 | 88.75 ± 22.49 |
Why the gap arises there. SC extrapolation has the longest horizon in our benchmark: four held-out time points, up to day 18, each predicted by integrating from . A deviation at the end of the training range is therefore carried into the extrapolation and grows with the horizon. How strongly it grows depends on the model, and despite the shared VAE architecture, scNODE and our first-order ablation differ in most implementation choices, including VAE training (joint vs. pretrained and frozen), the space of the loss (gene vs. latent) and the ODE solver (Euler with step size 1 vs. adaptive dopri5).
Decoder or latent dynamics? As a diagnostic, we replaced the linear decoder of scNODE with a nonlinear one (one hidden layer of width 50 and a ReLU output), keeping all other settings. This lowers the SC extrapolation to 62.43 ± 9.78. Its spread across seeds, however, remains large (s.d. 9.78, 4.7 times that of the first-order ablation), so the instability originates in the latent dynamics of scNODE beyond the last observed time point, which a decoder cannot correct.
We sincerely thank the reviewer for the constructive comments. In our updated manuscript, we highlight the changes in blue.
Weakness 1
We have added Cell-MNN (von Bassewitz et al., 2026) and LGP-OT (Balik & Lähdesmäki, 2026) to the Related Work and as baselines to Tables 1 and 2, for both interpolation and extrapolation. Both were run from their official implementations with the settings recommended by their authors, without tuning, on the splits and seeds and with the evaluation pipeline of Tables 1 and 2 (Table R3-1).
Table R3-1. Temporal prediction results of Cell-MNN and LGP-OT; mean ± s.d. over five seeds.
scMomenta achieves the lowest W2 of the three methods on all five datasets, in both interpolation and extrapolation. It also has the highest Pearson in eight of the ten entries; the two exceptions are interpolation on SC, where Cell-MNN is higher, and on DR, where both methods are higher. LGP-OT is a strong competitor: its W2 is within 5% of that of scMomenta on DR in both tasks and on GA extrapolation. Cell-MNN was evaluated only on interpolation in its paper; we obtained its extrapolations by integrating the learned dynamics beyond the last observed time point. It predicts with the closed-form solution exp(AθΔt)z of a linear ODE fixed at the starting state, which its authors use as a local approximation, so any error in Aθ is amplified exponentially as the target time moves beyond the observed range.