- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper addresses the challenging task of extrapolating future single-cell gene-expression distributions and population composition from unpaired temporal snapshots. It proposes scHOMURA, a generative framework that directly learns transitions across multiple forecast horizons, including non-adjacent snapshots, and combines multi-horizon predictions with partial recursion at inference to reduce reliance on repeated one-step forecasting. The method first clusters cells in a shared latent space and uses prototype-level unbalanced optimal transport to construct cell-level pseudo-pairs, reducing the computational cost of establishing correspondences. It then jointly learns deterministic mean displacements, stochastic residuals through conditional flow matching, and a state-dependent growth field that adjusts the relative contributions of source cells. Distribution-matching, cross-horizon consistency, and growth-supervision losses further support prediction of future distributions and population composition. Experiments across multiple time-resolved single-cell datasets show strong forecasting performance, particularly in Wasserstein distance, alongside favorable results in extended-horizon settings compared with the evaluated baselines. These results are encouraging for the challenging problem of forecasting cellular populations beyond the observed temporal window.
Critical strengths
-
Clear methodological design combining multi-horizon transition learning with cellular heterogeneity modeling.
The method constructs pseudo-pairs using prototype-level UOT across multiple forecast horizons, including non-adjacent snapshots, reducing reliance on transitions between adjacent time points. It decomposes future state changes into deterministic mean displacements and stochastic residuals generated through conditional flow matching, capturing shared progression and diverse outcomes conditional on the source state, respectively. This decomposition provides a clear framework for jointly modeling population evolution and cellular heterogeneity. -
Comprehensive experiments examining predictive capabilities and component contributions.
The authors evaluate the method on five single-cell datasets with seven forecasting splits, assessing overall distributions, gene-expression changes, and population composition. Additional experiments investigate extended-horizon forecasting, rare-population expansion, and unseen-state recovery. The method achieves the lowest mean Wasserstein distance across all reported splits. Ablation studies further examine how the main components contribute across different settings and metrics, helping clarify the method’s benefits and limitations.
Critical weaknesses
-
The model predicts relative population composition but not absolute cell numbers. Although the method uses unbalanced optimal transport and a growth field, the latter operates through normalized source-contribution weights rather than explicitly modeling mechanistic proliferation and death. Consequently, the model cannot distinguish changes in absolute population size from changes in relative composition. This limits its applicability to biological questions where absolute cell counts and population expansion or depletion are important outcomes, beyond the relative proportions of different cellular states.
-
The ablation studies do not sufficiently isolate the sources of the reported improvements. On CD-long, removing multi-horizon learning barely changes the mean Wasserstein distance (16.35 versus 16.36), while removing growth modeling improves the mean values of all seven reported metrics, including Type-JS (0.079 to 0.061). In contrast, multi-horizon learning substantially improves abundance calibration on CD-short, increasing GrowthSkill from −1.329 to +0.172. These results suggest setting-dependent benefits rather than uniformly beneficial component contributions.
Questions for authors
-
Why use UOT rather than balanced OT? In Appendix A.1.1, both prototype mass vectors are normalized, satisfying , so balanced OT is also applicable. Could the authors explain this choice and provide an experimental comparison with balanced OT?
-
How are different forecast horizons sampled during training? For , are horizons sampled uniformly, are all eligible source–target snapshot pairs sampled uniformly, or is another scheme used? Please specify the proportion of training samples allocated to each horizon.
-
How does the training cost compare with other methods? Although the FM regression loss is simulation-free, the MMD and consistency objectives require numerical integration. Please provide comparisons of training time and peak memory usage with other algorithms under comparable settings.
-
Would consistency across different source snapshots improve forecasting? The current loss compares a direct two-step forecast with two composed one-step forecasts. Have the authors considered additionally matching the forecast distributions for the same target time from two observed source snapshots?
\left( \widehat{Z}^{\mathrm{direct}}_{t\rightarrow t+2}, \widehat{Z}^{\mathrm{direct}}_{t+1\rightarrow t+2} \right)$$ Since inference mixes forecasts from different source times, would this constraint offer benefits beyond the existing distribution-matching and consistency losses?
8 comments
Setting-dependent contributions of model components
Response to Weakness2
Thank you for this insightful comment. We agree that the original ablation results show setting-dependent rather than uniformly positive contributions from each component. To clarify these findings, we conducted additional five-seed experiments across different levels of temporal supervision and targeted rare/emerging-population benchmarks. The results reveal complementary roles: multi-horizon learning provides clearer benefits with richer temporal supervision, residual dynamics facilitate the recovery of emerging states, and growth modeling improves relative abundance calibration.
Temporal supervision and multi-horizon learning. CD-long contains only three training snapshots (E9.5–E11.5), providing two one-step transitions, one two-step pair, and no three-step pairs. This limits the multi-step supervision available to the model and may also make long-range growth extrapolation less reliable, potentially explaining why removing growth modeling improves the aggregate metrics on CD-long (Table 4). To investigate the influence of temporal supervision, we introduced CD-medium with five training snapshots and compared it with CD-short, which has seven. As shown in Table R14-1, the full model achieves better aggregate performance across all seven metrics on CD-medium and CD-short, while providing only modest improvements on CD-long.
Table R14-1. Effect of multi-horizon learning under different levels of temporal supervision.
| Split | Methods | W₂ ↓ | DE-AP ↑ | DE-dir ↑ | Δr ↑ | MMD ↓ | Variogram ↓ | Type JS ↓ |
|---|---|---|---|---|---|---|---|---|
| CD-long | Full | 16.348 ± 0.169 | 0.298 ± 0.039 | 0.768 ± 0.035 | 0.369 ± 0.041 | 0.132 ± 0.015 | 0.023 ± 0.001 | 0.079 ± 0.006 |
| w/o multi-horizon | 16.358 ± 0.157 | 0.306 ± 0.015 | 0.768 ± 0.073 | 0.365 ± 0.069 | 0.141 ± 0.011 | 0.024 ± 0.003 | 0.083 ± 0.013 | |
| CD-medium | Full | 17.163 ± 0.216 | 0.297 ± 0.056 | 0.610 ± 0.066 | 0.130 ± 0.131 | 0.123 ± 0.007 | 0.024 ± 0.001 | 0.171 ± 0.021 |
| w/o multi-horizon | 17.426 ± 0.302 | 0.258 ± 0.029 | 0.545 ± 0.032 | 0.049 ± 0.057 | 0.148 ± 0.011 | 0.027 ± 0.004 | 0.197 ± 0.013 | |
| CD-short | Full | 15.091 ± 0.157 | 0.263 ± 0.085 | 0.575 ± 0.118 | 0.261 ± 0.059 | 0.044 ± 0.008 | 0.014 ± 0.001 | 0.013 ± 0.005 |
| w/o multi-horizon | 15.482 ± 0.344 | 0.235 ± 0.063 | 0.478 ± 0.084 | 0.181 ± 0.103 | 0.048 ± 0.009 | 0.016 ± 0.002 | 0.022 ± 0.012 |
To further assess whether these improvements are consistent across random seeds, we computed paired differences using five matched seeds (Table R14-2). The mean improvement increases from 0.010 on CD-long to 0.262 on CD-medium and 0.391 on CD-short. Importantly, on CR-long, the full model outperforms the ablation in all five seeds, with a mean improvement of 0.276. These results provide stronger evidence for the setting-dependent benefits of multi-horizon learning, while confirming its limited effect on CD-long.
**Table R14-2. Paired-seed analysis of multi-horizon learning across forecasting splits.**Paired improvement is defined as . Confidence intervals are nominal 95% paired intervals.
| Split | Full W₂ ↓ | w/o multi-horizon W₂ ↓ | Paired improvement ↑ | 95% paired CI | Full better |
|---|---|---|---|---|---|
| CD-long | 16.348 ± 0.169 | 16.358 ± 0.157 | +0.010 | [−0.178, +0.199] | 3/5 |
| CD-medium | 17.163 ± 0.216 | 17.426 ± 0.302 | +0.262 | [+0.009, +0.515] | 4/5 |
| CD-short | 15.091 ± 0.157 | 15.482 ± 0.344 | +0.391 | [−0.006, +0.787] | 4/5 |
| CR-long | 17.796 ± 0.242 | 18.072 ± 0.169 | +0.276 | [+0.054, +0.497] | 5/5 |
Rare-population abundance calibration. Multi-horizon learning also improves the calibration of rare-population expansion on CD-short (Table R14-3). Although removing multi-horizon learning improves expansion ranking and direction accuracy, the full model nearly halves abundance-change error and improves GrowthSkill from negative to positive. The immune-cell predictions provide a concrete example (Table R14-4): both variants anticipate an increasing trend, but the single-horizon model substantially overestimates future immune-cell abundance, whereas the full model produces predictions closer to the observed proportions. These results highlight the distinction between identifying expanding populations and accurately predicting their expansion magnitudes.
Table R14-3. Effect of multi-horizon learning on rare-population expansion in CD-short.
| Methods | Expn. AP ↑ | Dir. ↑ | ΔMAE ↓ | GrowthSkill ↑ |
|---|---|---|---|---|
| Full | 0.800 ± 0.209 | 0.650 ± 0.137 | 0.015 ± 0.003 | 0.172 ± 0.251 |
| w/o multi-horizon | 0.967 ± 0.075 | 0.750 ± 0.000 | 0.029 ± 0.012 | −0.667 ± 0.723 |
The remaining results of this response are provided in the next part due to the word limit.
Setting-dependent contributions of model components(Part 2)
Response to Weakness2(Part 2)
Table R14-4. Comparison of predicted and observed immune-cell proportions at future developmental stages in CD-short.
| Target time | True immune-cell proportion | Full | w/o multi-horizon |
|---|---|---|---|
| E16.5 | 4.19% | 5.80% | 7.69% |
| E17.5 | 7.42% | 8.26% | 12.27% |
Contributions of residual dynamics and growth modeling. To further distinguish their roles, we evaluated the ablated variants directly on rare-state expansion and natural emergence benchmarks (Table R14-5). On CD-short, removing growth modeling substantially degrades expansion direction accuracy and abundance calibration. In contrast, removing residual flow has relatively minor effects on abundance calibration but eliminates accepted emerging-state predictions on BetaDiff. Removing growth modeling also reduces accepted emerging-cell mass from 0.064 to 0.023. These findings suggest that residual dynamics are particularly important for reaching novel cellular states, whereas growth modeling regulates their relative representation in the predicted population.
Table R14-5. Contributions of residual dynamics and growth modeling to rare-state expansion and natural emergence. The first four metrics evaluate rare-state expansion on CD-short, while the last three evaluate natural emergence on BetaDiff.
| Methods | Expn. AP ↑ | Dir. ↑ | ΔMAE ↓ | GrowthSkill ↑ | Accepted ↑ | Unsupported ↓ | Precision ↑ |
|---|---|---|---|---|---|---|---|
| Full | 0.800 ± 0.209 | 0.650 ± 0.137 | 0.015 ± 0.003 | 0.172 ± 0.251 | 0.064 ± 0.028 | 0.043 ± 0.017 | 0.594 ± 0.047 |
| w/o residual flow | 0.883 ± 0.183 | 0.700 ± 0.112 | 0.016 ± 0.005 | 0.107 ± 0.272 | 0.000 ± 0.000 | 0.000 ± 0.000 | — |
| w/o growth modeling | 0.617 ± 0.126 | 0.250 ± 0.000 | 0.027 ± 0.002 | −0.511 ± 0.187 | 0.023 ± 0.013 | 0.014 ± 0.010 | 0.636 ± 0.101 |
| Persistence | 0.633 ± 0.209 | 0.450 ± 0.112 | 0.018 ± 0.001 | 0.000 ± 0.000 | 0.041 ± 0.001 | 0.055 ± 0.003 | 0.427 ± 0.010 |
| Noise-halo | 0.683 ± 0.137 | 0.450 ± 0.112 | 0.018 ± 0.002 | −0.001 ± 0.006 | 0.101 ± 0.006 | 0.107 ± 0.008 | 0.485 ± 0.029 |
Controlled masked emergence (BetaDiff). The masked-emergence experiments further illustrate the complementary contributions of these components (Tables R14-6 and R14-7). For the rare sst_hhex population, removing residual flow eliminates detection entirely, while removing growth modeling reduces detection to 2/10 and predicted abundance from 2.20% to 0.30%. For sc_alpha, which follows a more densely observed developmental trajectory, both ablated variants still permit frequent detection. However, removing growth modeling substantially reduces predicted abundance. These results indicate that residual dynamics are particularly important for recovering poorly supported future states, while growth modeling improves their representation even when the target state is already reachable.
Table R14-6. Contributions of residual dynamics and growth modeling to masked emergence recovery of the rare sst_hhex population on BetaDiff.
| Methods | Detected ↑ | Predicted Fraction | Conditional W₂ ↓ | OOS-precision ↑ |
|---|---|---|---|---|
| Full | 7/10 | 2.20% | 16.3 | 0.036 |
| w/o residual flow | 0/10 | 0.00% | — | — |
| w/o growth modeling | 2/10 | 0.30% | 17.5 | 0.022 |
| Persistence | 0/10 | 0.10% | — | — |
| Noise-halo | 0/10 | 0.10% | — | 0 |
Table R14-7. Contributions of residual dynamics and growth modeling to masked emergence recovery of the sc_alpha population on BetaDiff.
| Methods | Detected ↑ | Predicted Fraction | Conditional W₂ ↓ | OOS-precision ↑ |
|---|---|---|---|---|
| Full | 10/10 | 12.10% | 23.3 | 0.477 |
| w/o residual flow | 8/10 | 5.60% | 22.6 | — |
| w/o growth modeling | 10/10 | 3.70% | 23.5 | 0.724 |
| Persistence | 1/10 | 0.40% | 23.7 | — |
| Noise-halo | 10/10 | 2.90% | 27.3 | 0.071 |
Together, these targeted analyses clarify why the components show setting-dependent contributions rather than uniformly improving every aggregate metric. Multi-horizon learning provides clearer benefits under richer temporal supervision and improves rare-population abundance calibration. Conditional residual dynamics facilitate recovery of novel or poorly supported future states, whereas the growth field adjusts their relative population contributions, with potential limitations under sparse temporal supervision. We have incorporated these results into the new full PDF.
Relative versus absolute population dynamics
Response to Weakness1
Thank you for this insightful comment. We agree that distinguishing absolute population expansion or depletion from changes in relative composition is biologically important. However, inferring absolute population sizes from standard unpaired scRNA-seq snapshots faces a fundamental limitation: the total number of cells in the underlying biological system is generally not observed.
In standard scRNA-seq experiments, the number of profiled cells at each time point depends on experimental sampling, dissociation efficiency, and capture efficiency, rather than directly reflecting the total biological population size. For example, a developing tissue may undergo substantial proliferation without a corresponding increase in the number of sequenced cells. Without external measurements of total cell counts or calibrated sampling fractions, absolute population growth cannot be reliably distinguished from changes in sampling scale. Such information is unavailable in the public benchmarks considered in our study.
Accordingly, scHOMURA is designed to predict relative population composition rather than absolute cell numbers. Its state-dependent growth field adjusts the relative contributions of source cellular states to future distributions, capturing differential changes in population representation without explicitly estimating mechanistic proliferation or death rates. This distinction is important: the reported rare-population expansion metrics quantify changes in relative abundance, not absolute increases in cell numbers. Nevertheless, forecasting relative composition remains biologically informative for identifying populations that become more or less prevalent during development and differentiation.
We acknowledge that extending scHOMURA to absolute population dynamics would require additional information beyond uncalibrated scRNA-seq snapshots. Incorporating independent cell-count measurements or calibrated sampling fractions, together with explicit models of proliferation and death, represents an important direction for future work.