- Rating
- 3
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper proposes OmniMorph, reportedly the first unified generative model for Cell Painting, built on flow matching with a DiT backbone and disentangling channel semantics from task role via channel and role embeddings. It is trained jointly in two stages on six public datasets and four tasks (morphology generation, perturbation prediction, missing-channel completion, brightfield-to-fluorescence), conditioned on Uni-Mol/ESM-2 representations and Qwen2.5-14B-encoded instruction text. It leads in most settings and ranks first on all eight unseen-perturbation (OOD) morphology-generation settings.
Critical strengths
-
Clear and elegant problem formulation: the paper identifies heterogeneity in channel count and labeled subcellular structures as the central obstacle, and uniformly formalizes the four tasks as "experimental condition + role assignment". The formalization is clean and leaves room for extension.
-
Unusually rigorous evaluation: frozen target lists and biology panels, ID/OOD splits with a zero-overlap audit, and all baselines retrained by the authors under identical splits and training exposure (Table 20) rather than quoting published numbers.
-
Evaluation beyond image-distribution metrics: beyond FID/KID, MoA/perturbation balanced accuracy is measured with classifiers trained on real images only, plus effect-direction and dose-response analyses. The evaluation spans distribution, discrimination, and biological direction.
-
Thorough and honest ablations: removing the channel embedding raises the mean FID from 50.4 to 152.9 (Table 5a), directly supporting the core design. The text ablation (Table 10) honestly reveals that the caption carries dataset identity.
-
Strong generalization to unseen perturbations: OmniMorph ranks first on all eight OOD morphology-generation settings (Table 1), whereas MorphGen and MDTv2 rely on a closed perturbation vocabulary. This is the strongest evidence for cross-dataset knowledge sharing.
Critical weaknesses
-
Limited methodological novelty, with the text-condition framing overstated: every component is existing technology, and the core novelty—disentangling channel and role embeddings—is closer to systematization than a breakthrough. The claimed "natural-language instruction" is actually template field filling (Table 15), weakening the persuasiveness of the method section.
-
Fairness concerns in the comparison: OmniMorph's total budget is (200k+200k)×8 while each baseline gets only 200k×8 (Table 20), giving the unified model twice the exposure and all six datasets. In perturbation prediction, IMPA is the only baseline actually conditioned on a control image, and CellFlux (ICML 2025), cited as representative, is not compared.
-
Substantive gaps in biological fidelity: dose response is recovered only for the strongest responders (e.g., 3 of 16 RxRx3 compounds reach a trajectory cosine above 0.3). Despite motivating the work as replacing part of screening experiments, no experiment demonstrates the downstream utility of the generated data.
Questions for authors
-
Could you report each baseline at 400k steps (matching the total budget), or OmniMorph trained for only 200k steps? On which metric and which validation set were the reported checkpoints selected?
-
Table 8 shows joint training clearly worse than per-dataset specialists on several FID settings (e.g., CPG0001 OOD 164.4 vs 25.5)—why is this degradation not discussed? Which model version does Table 5(b) actually use, given its numbers match Table 2 while the text says first-stage?
-
If the semantic channel embedding were replaced by a simple dataset-ID embedding (one learnable vector per dataset), how would the metrics change? This is the key control separating channel-semantic sharing from caption-carried dataset identity.
12 comments
Response to W2 (1/3): training budget and comparison with CellFlux
W2. Fairness concerns in the comparison: OmniMorph's total budget is (200k+200k)x8 while each baseline gets only 200k x8 (Table 20), giving the unified model twice the exposure and all six datasets. In perturbation prediction, IMPA is the only baseline actually conditioned on a control image, and CellFlux (ICML 2025), cited as representative, is not compared.
We thank the reviewer. Joint training does not give OmniMorph more exposure per dataset, and we added CellFlux. Under the budgets of Table 20, OmniMorph sees each training image 2 to 40 times less often than every baseline in all four tasks (Table R2), its first stage alone, without multi-task training, already outperforms the baselines (Tables R1 and R1b), and it is ahead of CellFlux and IMPA in 13 of the 16 perturbation-prediction settings (Table R8).
Data exposure. Each baseline spends its whole budget on the data covered by one set of its weights (Table 20), whereas OmniMorph shares its budget among six datasets and four tasks (Section 3). Table R2 divides the training samples of each method by the number of training images it covers (Table 16). While preparing it, we found a typo in the Models column of Table 20, which the revision corrects.
(a) Morphology generation
| BBBC021 | CPG0-C | CPG0-G | CPG0001 | CPG0004 | CPG0036 | RxRx3-C | RxRx3-G | |
|---|---|---|---|---|---|---|---|---|
| SD / MorphoDiff / MorphGen / MorphGen* | 215 | 77 | 104 | 88 | 82 | 170 | 18 | 12 |
| Mol2Image | 215 | 77 | – | 88 | 82 | 170 | 18 | – |
| MDTv2 / MDTv2* | 215 | 60 | 60 | 88 | 82 | 170 | 9.2 | 9.2 |
| OmniMorph | 19.6 | 10.4 | 10.4 | 12.6 | 12.1 | 17.5 | 4.1 | 4.1 |
| OmniMorph, first stage | 17.9 | 9.4 | 9.4 | 11.4 | 11.0 | 15.9 | 3.7 | 3.7 |
(b) Channel completion
| BBBC021 | CPG0000 | CPG0001 | CPG0004 | CPG0036 | RxRx3 | |
|---|---|---|---|---|---|---|
| U-Net / Pix2Pix / MCD | – | 24.8 | 24.8 | 24.8 | – | 9.2 |
| U-Net / Pix2Pix (validation plate held out) | – | 28.0 | 28.0 | 28.0 | – | 9.3 |
| OmniMorph | 5.4 | 2.8 | 3.4 | 3.3 | 4.8 | 1.1 |
(c) Brightfield-to-fluorescence, per training pair (CPG0000 and CPG0001)
| CPG0000 and CPG0001 | |
|---|---|
| cWGAN-GP / ISL / CellResDM | 43.0 |
| cWGAN-GP / ISL (validation plate held out) | 50.4 |
| OmniMorph | 12.9 |
(d) Perturbation prediction
| BBBC021 | CPG0-C | CPG0-G | CPG0001 | CPG0004 | CPG0036 | RxRx3-C | RxRx3-G | |
|---|---|---|---|---|---|---|---|---|
| IMPA / CellFlux† | 215 | 60 | 60 | 88 | 82 | 170 | 9.2 | 9.2 |
| OmniMorph | 5.4 | 2.8 | 2.8 | 3.4 | 3.3 | 4.8 | 1.1 | 1.1 |
Table R2. Average number of times each training image is seen at the budgets of Table 20. A view counts its perturbation type and all control images of the dataset, which gives the baselines the lowest repetition. Where one set of weights covers several columns (scope "source" or "channel count" in Table 20), its value is repeated in each of them. "–" means not covered. First stage is the first-stage morphology-generation training alone. *Retrained with OmniMorph's frozen condition encoders. †Trained on the SD-VAE latents shared by all latent-space methods, with the same frozen Uni-Mol / ESM-2 embeddings as IMPA.
Response to W2 (2/3): training budget and comparison with CellFlux
W2 (continued).
Matched data and budget. The first-stage models train morphology generation only, with no more than 200k×8 samples (Table 20). Trained on one dataset at a time, on the same data as a single-dataset baseline, the per-dataset model is ahead of every baseline in 11 of the 16 settings, including 6 of the 8 OOD settings. Since MorphGen* and MDTv2* use the same frozen encoders, this gain comes from the architecture. Trained on all six datasets within the same budget, the multi-dataset model sees each training image 2 to 12 times less often than a baseline and is ahead of every baseline of the original submission in 11 of the 16 settings (Table R1). These two first-stage models are the two rows of Table 8, whose multi-dataset row we correct in our answer to Q2.
| Method | BBBC021 ID | CPG1 ID | CPG4 ID | CPG36 ID | CPG0-C ID | CPG0-G ID | RxRx3-C ID | RxRx3-G ID | BBBC021 OOD | CPG1 OOD | CPG4 OOD | CPG36 OOD | CPG0-C OOD | CPG0-G OOD | RxRx3-C OOD | RxRx3-G OOD |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Stable Diffusion | 214.9 | 141.0 | 193.7 | 264.8 | 218.1 | 200.1 | 108.7 | 197.7 | 235.6 | 150.9 | 221.5 | 260.0 | 222.9 | 195.9 | 112.6 | 181.9 |
| Mol2Image | 139.0 | 216.9 | 60.8 | 138.0 | 119.9 | – | 190.4 | – | 128.1 | 211.0 | 52.7 | 142.0 | 111.6 | – | 186.3 | – |
| MorphoDiff | 142.2 | 136.3 | 128.2 | 277.8 | 202.3 | 192.3 | 116.5 | 209.2 | 180.1 | 117.3 | 150.0 | 272.0 | 208.0 | 190.3 | 120.7 | 192.6 |
| MorphGen | 60.5 | 49.4 | 47.7 | 82.2 | 110.1 | 106.2 | 40.0 | 75.8 | – | – | – | – | – | – | – | – |
| MorphGen* | 52.3 | 48.9 | 41.5 | 86.9 | 38.1 | 61.6 | 28.7 | 65.6 | 42.3 | 42.1 | 33.7 | 87.7 | 34.9 | 51.0 | 29.1 | 62.5 |
| MDTv2 | 46.9 | 74.8 | 26.5 | 104.7 | 71.5 | 60.3 | 53.1 | 46.1 | – | – | – | – | – | – | – | – |
| MDTv2* | 60.4 | 101.0 | 32.7 | 118.1 | 41.0 | 62.9 | 102.6 | 90.5 | 42.2 | 87.6 | 33.4 | 111.6 | 42.9 | 53.7 | 100.3 | 89.4 |
| OmniMorph (Table 1) | 43.9 | 50.4 | 32.7 | 93.5 | 52.9 | 61.1 | 29.5 | 25.7 | 32.3 | 90.0 | 27.0 | 93.1 | 51.2 | 50.9 | 29.1 | 26.3 |
| OmniMorph, per-dataset (first stage) | 43.8 | 38.1 | 43.0 | 99.0 | 36.5 | 56.4 | 29.9 | 29.2 | 27.3 | 25.5 | 29.7 | 100.2 | 32.6 | 45.7 | 31.1 | 28.1 |
| OmniMorph, multi-dataset (first stage) | 47.7 | 51.3 | 38.3 | 102.2 | 56.5 | 56.2 | 30.7 | 30.7 | 35.3 | 118.0 | 30.4 | 100.3 | 48.2 | 46.0 | 31.3 | 29.9 |
Table R1. Morphology-generation FID (↓) on the targets of Table 1: the revised Table 1 with the two first-stage models added. *With OmniMorph's frozen open encoders. "–": a closed-vocabulary model cannot generate unseen perturbations. Bold: best of each column; the counts in the text compare each first-stage row with the baselines only.
Pretraining and encoders. Neither first-stage model has multi-task training, and the per-dataset model sees no other dataset. The benefit of pretraining for the other three tasks is measured in Table 5(c), where the single-task model trained from scratch is the weakest on every metric. Frozen condition encoders are also used by IMPA and CellFlux† (Uni-Mol / ESM-2) and by MorphGen* and MDTv2* (those of OmniMorph).
Biological fidelity. Without a dataset identifier, the effect of a compound is learned from every dataset in which it occurs. On the classification panels of Table 2, the multi-dataset first-stage model is ahead of every baseline on four of the five panels (Table R1b) and ahead of the per-dataset models on four of the five (Table 5(b)).
| BBBC021 (MoA, 6) | CPG0001 (MoA, 10) | CPG0004 (cmpd, 10) | RxRx3-C (cmpd, 15) | RxRx3-G (gene, 15) | |
|---|---|---|---|---|---|
| Real images (ref.) | 0.954 | 0.815 | 0.571 | 0.620 | 0.760 |
| Stable Diffusion | 0.157 | 0.088 | 0.083 | 0.065 | 0.081 |
| Mol2Image | 0.262 | 0.095 | 0.195 | 0.099 | – |
| MorphoDiff | 0.253 | 0.092 | 0.108 | 0.087 | 0.085 |
| MDTv2 | 0.250 | 0.299 | 0.175 | 0.237 | 0.242 |
| MDTv2* | 0.206 | 0.206 | 0.134 | 0.108 | 0.107 |
| MorphGen | 0.670 | 0.532 | 0.236 | 0.366 | 0.304 |
| MorphGen* | 0.576 | 0.462 | 0.217 | 0.372 | 0.148 |
| OmniMorph (Table 2) | 0.703 | 0.649 | 0.247 | 0.476 | 0.526 |
| OmniMorph, multi-dataset (first stage) | 0.642 | 0.633 | 0.237 | 0.473 | 0.460 |
Table R1b. Morphology generation, balanced accuracy (↑) of the classifier trained on real images only (Table 2), on RGB renderings. *Retrained with OmniMorph's frozen open encoders. Bold: best generated entry of each column.
Response to W1: novelty and the text condition
We agree that the individual components exist and that the text is templated, and the revision says so. Our contribution is the formulation that lets one model cover heterogeneous panels and tasks, and the templated text carries real conditioning signal.
Novelty. As stated in the Introduction, our contribution is that, for one model to cover several datasets and tasks, we disentangle the semantics of a channel from its role in the task with two embeddings added to all tokens of the channel. Public Cell Painting datasets differ in their panels and in whether brightfield images and controls are acquired, so a generator for all of them has to accept varying channel combinations and several tasks. With the channel tokens arranged along the sequence and tagged by the structure they stain, a change of panel becomes a change of sequence length, and a channel is identified by what it stains rather than by its position in a dataset. The role embedding then marks each channel as an observed channel, a brightfield input, a control reference or a generation target, so that the four tasks share one model. As the reviewer notes, "the formalization is clean and leaves room for extension".
What the formulation enables. Both embeddings are needed. The channel embedding lowers the mean FID of morphology generation from 152.9 to 50.4 (Table 5(a)), and removing the role embedding raises FID in 10 of the 12 morphology-generation settings, by up to 19.8 (new role-embedding ablation in the appendix). The formulation also gives capabilities that separate models do not have. BBBC021 has only DNA, tubulin and actin, and OmniMorph generates RNA, ER and mitochondria channels for it, learned from the other datasets (new appendix subsection). On JUMP plates unseen in training, the jointly trained OmniMorph separates the compounds best, reaching 85% of the real-image ceiling (new appendix subsection).
Text condition. We agree that the text is produced by templates (Table 15) and that "natural-language instruction" overstates it. The revision calls it a templated text instruction (abstract, Introduction and Section 3). The templates are a deliberate choice. They give every dataset the same format, so that the same field means the same thing across datasets. The text is encoded by a frozen language model (Qwen2.5-14B), so the format can later be extended to free text without changing the model.
The templated text carries real conditioning signal. We trained a model in which the four structured conditions (perturbation type, perturbation identity, dose and cell line) are removed while the text and the role embedding are kept, and compared it with the full-condition model. Both are initialised from the first-stage model and then trained jointly on the four tasks for 100k steps at an effective batch of 8, half of the 200k×8 joint training of the final model (Table 20), so only the comparison between the two is meaningful. On the text-only generations, the frozen real-image classifier of Table 2 identifies the MoA or perturbation above chance on every panel, at 2.5 to 5 times chance on BBBC021, CPG0001 and RxRx3 compounds (Table R17).
Table R17. Balanced accuracy (↑) of the frozen real-image classifier of Table 2 on the generations of the text-only model.
Conversely, the text-condition ablation of Table 10 removes the description of the sample and keeps the task instruction and the structured conditions. The Pearson correlation of channel completion then drops by 0.04 to 0.08 on all four datasets, and morphology generation collapses on CPG0000 and CPG0036, with FID rising from 51–61 to 116–168 and from 93 to over 200. Since the text alone carries perturbation information, a perturbation type for which no dedicated encoder exists can also enter through its text description, without attaching a further foundation model such as ESM-2 or Uni-Mol.
In summary, the components are known, and our contribution is the formulation that disentangles channel semantics from task roles, which the ablations support and which enables channel generation across datasets and transfer to unseen plates. The text condition is templated, as the revision now states, and it carries real conditioning signal.