- Rating
- 3
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper proposes SDFM, a knowledge-conditioned flow-matching framework for predicting single-cell genetic perturbation responses. It incorporates perturbation representations derived from STRING and Gene Ontology, multi-head cross-attention conditioning, an interaction-aware composer for multi-gene perturbations, and mini-batch optimal-transport pairing. Beyond local velocity matching, its training objective adds one-step endpoint-based population supervision and differentiable rollout losses. The work validates the method on single-gene and combinatorial perturbation benchmarks with component ablation studies.
Critical strengths
-
Strong benchmark improvements with substantial gains across most evaluation metrics.
-
Good reproducibility. Training configurations are well documented, and the released code enables straightforward pipeline reproduction.
Critical weaknesses
-
Limited methodological novelty. The rationalization for integrating different components is insufficient; the main contribution lies in combining existing techniques.
-
Aggregate improvements mask the distribution of benefits from the knowledge graph. A subset of perturbations may see degraded performance, limiting the generalizability of the conclusions.
Questions for authors
This work presents a promising framework for single-cell genetic perturbation prediction and demonstrates meaningful benchmark gains, yet I have three main concerns regarding interpretation of the results:
-
Clarification on Teacher-Δ: The teacher endpoint uses . This formulation follows the ideal straight conditional path, yet the learned velocity field varies along the generated trajectory. Could the mismatch between this one-step extrapolation and the integrated endpoint introduce systematic bias?
-
Please report the distribution of gains from knowledge conditioning, including the fraction of test perturbations with improved/worsened predictions and per-perturbation gain statistics. This helps evaluate whether the reported average improvement transfers to genes with limited prior annotations.
-
How is the default top-20 edge setting selected? Are the main conclusions stable under sparser or denser graph constructions?
5 comments
Q2. Please report the distribution of gains from knowledge conditioning, including the fraction of test perturbations with improved or worsened predictions and per-perturbation gain statistics. This would help determine whether the benefits generalize to perturbations involving genes with limited prior annotations.
R2. We thank the reviewer for this insightful suggestion. We agree that aggregate performance alone may obscure the heterogeneous effects of biological knowledge conditioning across individual perturbations.
To address this concern, we analyzed the per-perturbation performance changes associated with knowledge conditioning across the three datasets. The results are summarized below:
| Dataset | Test Perturbations | Improved | Worsened | Mean Gain | Median Gain |
|---|---|---|---|---|---|
| Adamson | 22 | 18 (81.82%) | 4 (18.18%) | 0.10653 | 0.06145 |
| HepG2 | 594 | 376 (63.30%) | 218 (36.70%) | 0.04631 | 0.04100 |
| Norman | 107 | 55 (51.40%) | 52 (48.60%) | 0.03247 | 0.00580 |
The results show that knowledge conditioning improves predictions for the majority of test perturbations across all three datasets. The improvement is particularly pronounced on Adamson, where 81.82% of perturbations benefit from biological knowledge. On HepG2, 63.30% of perturbations show improvements, while Norman exhibits a more balanced distribution.
Importantly, we acknowledge that knowledge conditioning does not benefit every perturbation. In particular, approximately half of the Norman test perturbations show reduced performance, and its median gain is relatively small. This indicates that the effectiveness of biological priors depends on the perturbation and dataset, rather than providing uniform improvements.
We further refer to our existing gene-family holdout evaluation (Table 20), which examines generalization beyond gene families represented during training. SDFM achieves Pearson Δ of 0.5987, DE100 of 0.6324, and normalized retrieval of 0.7215, outperforming the evaluated comparison methods in overall response correlation and retrieval.
Together, these analyses provide a more detailed characterization of the benefits and limitations of knowledge conditioning. We will include the per-perturbation statistics and clarify that biological knowledge provides an overall benefit while its effectiveness remains perturbation-dependent.
Q3. How was the default top-20 edge setting selected? Are the main conclusions stable under sparser or denser graph constructions?
R3. We thank the reviewer for this helpful question. The top-20 setting was adopted as a practical graph sparsification choice to retain informative local gene neighborhoods while controlling graph density and computational complexity. We agree that its sensitivity should be evaluated rather than assuming it is uniquely optimal.
To examine this, we compared sparser (top-10) and denser (top-40) graph constructions with the default top-20 setting. The results are summarized below:
| Dataset | Graph | Pearson Δ | Normalized Retrieval |
|---|---|---|---|
| Adamson | Top-10 | 0.7696 ± 0.0041 | 0.6127 ± 0.0029 |
| Adamson | Top-20 | 0.8071 ± 0.0084 | 0.6649 ± 0.0081 |
| Adamson | Top-40 | 0.8105 ± 0.0083 | 0.6697 ± 0.0092 |
| HepG2 | Top-10 | 0.2138 ± 0.0036 | 0.5379 ± 0.0057 |
| HepG2 | Top-20 | 0.2634 ± 0.0032 | 0.5751 ± 0.0062 |
| HepG2 | Top-40 | 0.2593 ± 0.0056 | 0.5802 ± 0.0075 |
| Norman | Top-10 | 0.2840 ± 0.0058 | 0.5201 ± 0.0036 |
| Norman | Top-20 | 0.3266 ± 0.0043 | 0.5649 ± 0.0061 |
| Norman | Top-40 | 0.3271 ± 0.0062 | 0.5564 ± 0.0063 |
The results show that the top-40 configuration achieves performance comparable to top-20, with slightly different trade-offs across datasets and metrics. Specifically, top-40 achieves slightly higher Pearson Δ on Adamson and Norman, while top-20 performs better on HepG2. Regarding Normalized Retrieval, top-40 performs slightly better on Adamson and HepG2, whereas top-20 performs better on Norman.
In contrast, the sparser top-10 configuration consistently reduces both Pearson Δ and Normalized Retrieval across all three datasets, suggesting that excessive graph sparsification may remove useful biological connectivity.
Overall, these results indicate that SDFM is relatively stable when increasing graph density from 20 to 40 neighbors, while retaining sufficient graph connectivity is important for predictive performance. We will clarify the rationale for the default top-20 setting and report these sensitivity results in the revised manuscript.
Q1. The Teacher-Δ endpoint is obtained through a one-step extrapolation along the ideal straight conditional path. However, the learned velocity field may vary along the generated trajectory. Could the mismatch between this one-step approximation and the fully integrated ODE endpoint introduce systematic bias?
R1. We thank the reviewer for raising this important question. We agree that the one-step Teacher-Δ endpoint is an approximation and may differ from the endpoint obtained through full ODE integration when the learned velocity field varies along the trajectory.
Specifically, Teacher-Δ estimates the endpoint from an intermediate state as:
x~p=ρ(xt+(1−t)vθ(xt,t,xc,ep))
Under the ideal conditional velocity field, where the velocity is constant along the straight conditional path, this extrapolation recovers the exact endpoint. However, for a learned state- and time-dependent velocity field, the one-step estimate may introduce approximation errors.
Importantly, SDFM does not treat the Teacher-Δ endpoint as the actual generated endpoint. Instead, it serves as an auxiliary population-level supervision signal. The complete ODE trajectory is independently constrained through two additional objectives:
Therefore, the overall objective combines local velocity matching, approximate teacher endpoint supervision, and supervision based on the complete integrated trajectory.
We acknowledge that the one-step approximation may introduce local discrepancies, but the additional rollout-based objectives explicitly account for the learned velocity dynamics throughout the integration process. We will clarify this distinction in the revised manuscript to avoid implying that the one-step teacher endpoint is identical to the actual ODE-generated endpoint.