- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper addresses the problem of forecasting future single-cell transcriptomic trajectories for unseen (held-out) perturbations given only their earliest observed post-perturbation response and time-matched control populations across the forecast horizon. The proposed framework, DeFo, uses optimal transport to align perturbed cells with control trajectories and decomposes the velocity field into a control-driven base field () and a perturbation-specific response field () inferred via Temporal Response Memory Attention (TRMA), which retrieves and linearly combines finite-difference mean effect trajectories from training perturbations based on response similarity. Experiments on a 2D synthetic system and two real time-resolved perturbation datasets (HSPC-Mk and RENGE) evaluate pseudobulk expression recovery and differential-expression direction concordance.
Critical strengths
-
Formulating held-out temporal perturbation forecasting conditioned on early-response observations and time-matched controls addresses a practical experimental design scenario, and separating shared control progression from perturbation-specific deviations provides a clean control-referenced coordinate system (Lotfollahi et al., 2019).
-
The manuscript is clearly organized, evaluates five condition-level holdout splits with reported standard deviations, and achieves strong pseudobulk differential-expression direction agreement on the RENGE genetic perturbation dataset.
Critical weaknesses
-
Despite being framed as a continuous flow-matching dynamical system, the forecasting ODE in Equation (14), , is entirely open-loop and does not depend on the evolving perturbed cell state on its right-hand side. Because future control populations are already provided at inference time and is simply a piecewise-linear interpolation of finite differences of population means in Equation (6), integrating Equation (14) via fourth-order Runge–Kutta is mathematically nearly equivalent to directly adding a static convex combination of known perturbations' mean displacements to the observed target-time control state, without learning state-dependent cellular dynamics (Tong et al., 2020; Zhang et al., 2024a).
-
Both the TRMA memory construction and the quantitative evaluation collapse single-cell distributions into population averages (pseudobulk), discarding single-cell heterogeneity. Specifically, Equations (4)–(6) average across all cells within each known perturbation before storing memory velocities, and all three metrics in Appendix A.4 (MSE, Pearson-Δ, and DE Dir.) evaluate only pseudobulk mean vectors or marginal gene-wise rank tests, with no single-cell distributional metrics (such as Wasserstein distance , Energy Distance, or MMD) to verify whether subpopulation structure and cell-level variance are preserved (Yeo et al., 2021).
-
The empirical results in Tables 1 and 2 do not convincingly justify the proposed modules across datasets. On HSPC-Mk (Table 1), the naive unconditional baseline Pert—which completely ignores the held-out perturbation's early response and simply pools known training perturbations—achieves a higher Pearson-Δ () than DeFo () and virtually identical DE Dir. ( vs. ), indicating that performance on this dataset is driven by shared homogeneous drug shifts rather than TRMA retrieval; meanwhile, on RENGE (Table 2), removing the core decomposition (w/o Deco), temporal velocity (w/o Temp), or soft composition (w/o Comp) yields nearly identical scores (e.g., MSE of for both DeFo and w/o Deco, and Pearson-Δ of vs. , well within one standard deviation).
-
The pipeline relies heavily on chained deterministic optimal-transport couplings across both time points and control-versus-perturbation conditions in a fixed 50-dimensional PCA space, which can introduce pairing artifacts when perturbations alter proliferation/apoptosis rates or shift subpopulation proportions (Zhang et al., 2024b). Furthermore, computing time-invariant attention weights solely from the earliest snapshot assumes that initial transient deviations uniquely determine long-term trajectories, which fails when distinct genetic perturbations share generic early stress responses or exhibit delayed transcriptional cascades.
Questions for authors
see weaknesses
2 comments
W3: Module effectiveness and dataset dependence
Pearson- and DE Direction assess response patterns and directions but not whether predictions preserve held-out-condition identity. We therefore added PDS, whose random expectation is 0.5.
| Method | HSPC-Mk PDS ↑ | RENGE PDS ↑ |
|---|---|---|
| Ctrl | 0.5000 ± 0.0000 | 0.5600 ± 0.0435 |
| Pert | 0.5056 ± 0.0124 | 0.5633 ± 0.0545 |
| scNODE | 0.4333 ± 0.0639 | 0.5600 ± 0.0303 |
| scGPT | 0.7667 ± 0.1155 | 0.5667 ± 0.0717 |
| CellFlow | 0.6833 ± 0.1155 | 0.5667 ± 0.0645 |
| scDFM | 0.5056 ± 0.0663 | 0.6167 ± 0.0264 |
| Departures | 0.6889 ± 0.0456 | 0.5700 ± 0.0908 |
| DeFo | 0.7833 ± 0.0719 | 0.6333 ± 0.0486 |
Pert has near-random PDS on HSPC-Mk despite strong directional scores, indicating that averaging reference perturbations preserves shared directions without reliably identifying the held-out condition. DeFo achieves the highest PDS on both datasets.
The original w/o Deco is a conventional condition-aware field and does not match DeFo’s inputs. We therefore added information-matched w/o Add, which receives the same matched control state, TRMA response, and time as DeFo but predicts total velocity with a single field.
| Model | HSPC MSE ↓ | Pearson- ↑ | DE Dir. ↑ | RENGE MSE ↓ | Pearson- ↑ | DE Dir. ↑ |
|---|---|---|---|---|---|---|
| w/o Add | 0.0034 ± 0.0007 | 0.4308 ± 0.1164 | 0.7148 ± 0.0410 | 0.0019 ± 0.0010 | 0.2753 ± 0.1176 | 0.8242 ± 0.1282 |
| DeFo | 0.0017 ± 0.0009 | 0.5627 ± 0.0800 | 0.8021 ± 0.0290 | 0.0010 ± 0.0007 | 0.4667 ± 0.0574 | 0.9465 ± 0.0110 |
Under matched conditioning information, replacing the full decomposed formulation with a unified field reduces performance on both datasets, supporting the utility of the full decomposed formulation.
To explain the smaller effects of w/o Temp and w/o Comp on RENGE, we computed median across-condition/across-time effect correlations: 0.475/0.291 for HSPC-Mk and 0.211/0.457 for RENGE. HSPC-Mk’s lower temporal consistency makes temporal transfer more useful, whereas RENGE’s higher consistency makes static responses and top-1 retrieval more competitive. Thus, the benefits of temporal memory and soft composition depend on response structure.
Finally, the reported standard deviations are computed across five condition splits after within-split averaging and therefore reflect held-out-condition composition, not initialization uncertainty or significance intervals. To assess initialization sensitivity separately, we repeated DeFo with seeds 43 and 44 on the same splits:
| Dataset | Seed | MSE ↓ | Pearson- ↑ | DE Dir. ↑ |
|---|---|---|---|---|
| HSPC-Mk | 42 | 0.0017 ± 0.0009 | 0.5627 ± 0.0800 | 0.8021 ± 0.0290 |
| HSPC-Mk | 43 | 0.0018 ± 0.0007 | 0.5422 ± 0.0821 | 0.7843 ± 0.0338 |
| HSPC-Mk | 44 | 0.0018 ± 0.0006 | 0.5817 ± 0.1129 | 0.8069 ± 0.0351 |
| RENGE | 42 | 0.0010 ± 0.0007 | 0.4667 ± 0.0574 | 0.9465 ± 0.0110 |
| RENGE | 43 | 0.0010 ± 0.0007 | 0.4768 ± 0.0448 | 0.9569 ± 0.0265 |
| RENGE | 44 | 0.0010 ± 0.0007 | 0.4515 ± 0.0534 | 0.9435 ± 0.0382 |
These three runs suggest limited sensitivity to initialization on the evaluated splits. Cross-split standard-deviation overlap alone therefore does not establish equivalence between models.
Overall, we have clarified DeFo’s model scope and added targeted evaluations of distribution recovery, component effectiveness, initialization stability, and OT robustness. We have also stated the limitations of early-response retrieval and condition-level response memory.
We thank the reviewer for the comments on DeFo’s dynamical form, single-cell heterogeneity, module effectiveness, and the assumptions underlying OT matching and early-response retrieval. We address these points below. Major revisions in the updated manuscript are shown in blue.
W1: Open-loop dynamics and anchor-time conditioning
We agree that Equation (14) does not depend on the evolving prediction x^(t), so DeFo is not a state-feedback dynamical system. However, prediction remains cell-specific because each TRMA query is the cell-level control-referenced effect Δxi(t0) at the anchor time. Different cells can therefore obtain different response combinations, although these weights remain fixed during forecasting.
Fixed weights do not imply a single static displacement because the memory values are time-varying response-velocity trajectories that are combined with the corresponding control progression. Continuous-time integration accumulates these time-varying components but does not introduce dependence on the generated state. We have therefore revised the model positioning to a continuous-time, control-guided trajectory model rather than a general state-dependent flow model.
W2: Single-cell heterogeneity and distributional evaluation
TRMA uses condition-average response keys and values to reduce sampling noise and avoid artificial cell-to-cell correspondence across perturbations. Cell specificity is retained through the query, attention weights, and cell-specific control reference, allowing different cells to retrieve different response combinations.
To evaluate population-level distribution recovery directly, we added Energy Distance (ED) in PCA-50 space.
DeFo achieves the lowest ED on RENGE and outperforms most learned methods on HSPC-Mk, although it is weaker than CellFlow and Pert there. These results indicate that cell-specific retrieval with control references preserves meaningful population-level structure, while condition-level memory remains limited for subgroup-specific effects.
W4: OT pseudo-trajectories and early-response retrieval
OT-reference robustness
DeFo propagates a single anchor-time perturbation–control OT coupling along the aligned control trajectory. To test whether the results depend on a particular coupling, we added Random fixed, which uses a random correspondence fixed across time, and Control mean, which removes cell-level matching and uses the control-population mean.
Both alternatives remain functional, while anchor-time OT gives the strongest response-level results. This indicates that the conclusions are not driven by one accidental coupling and that OT provides an informative control reference. This sensitivity analysis does not establish biological lineage correspondence, but it shows robustness to the construction of the control reference.
Early retrieval and delayed responses
TRMA determines its composition weights from the anchor-time effect, but the retrieved memory values are time-varying response-velocity trajectories rather than static displacements. Thus, early effects select reference response programs whose later memory values can contain delayed responses.
We compared condition-pair similarity matrices between the anchor time and each future time point. Their Spearman correlations were 0.62, 0.65, and 0.81 for HSPC-Mk and 0.69, 0.60, and 0.51 for RENGE. These positive correlations indicate that the relative response structure among perturbations remains partly preserved over time, supporting early-effect-based retrieval.
The limitation is that perturbations with indistinguishable early effects but different unobserved late mechanisms cannot be reliably separated. DeFo can transfer delayed patterns represented in memory, but cannot infer entirely novel late mechanisms unsupported by the early observation.