- Rating
- 3(+1)
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper studies phenotype-conditioned molecular generation when the molecular target is not necessarily known at inference time. PhenTTP transfers knowledge from a pretrained target-conditioned generator to phenotype-driven generation and supports both P-only and P+T inference from a shared model. Training combines phenotype–target alignment, phenotype-conditioned generation from phenotype–molecule pairs, and dual-route distillation. The method is evaluated on L1000 and Tahoe under four OOD settings using phenotype consistency, molecular quality, and target compatibility metrics.
Critical strengths
-
The problem formulation is interesting and practically relevant. Treating target-conditioned knowledge as a prior for phenotype-only generation is a natural way to handle the case where the molecular target is unavailable at inference. The shared P-only/P+T model is also a useful feature when target information is available.
-
The P-only results are encouraging. PhenTTP performs well on the phenotype metrics and outperforms the phenotype-only baselines on Vina across all four L1000 splits. Table 3 also shows that removing KD hurts the target-side results much more than removing P–T alignment.
Critical weaknesses
-
The role of P+T is less clear. Adding the target improves relative to P-only, but P+T also restores the native target input of the pretrained generator. In Appendix Table 8, the original DrugGPT T-only model is often better than P+T on affinity, and its phenotype scores are also frequently better. I therefore do not see clearly what phenotype conditioning adds in the P+T setting.
-
There are several design choices that are not well justified by the current ablations. For example, is split into and , but there is no comparison with using only , only , or no split. Also, KD-only is already comparable to the full model in Table 3, and joint training is competitive with staged training. Although the Tahoe ablations show benefits from alignment and staged training in some settings, their additional value on L1000 remains unclear.
Questions for authors
-
Why is P+T sometimes worse than the original T-only model, even on phenotype metrics? Do the authors think this comes from a phenotype–target trade-off, interference from the phenotype pathway, or limited joint supervision?
-
Do and provide distinct contributions in practice? I would like to see an ablation of the two pathways to understand whether the split is necessary.
-
How is PhenTTP implemented on ProtoBind-Diff? The DrugGPT implementation is described in much more detail. Are the intervention points and trainable modules the same across the two backbones?
-
Does P+T perform worse than T-only for most targets, or is the overall gap mainly driven by a few targets? I am also curious whether this difference is related to how much target-annotated training data is available for different targets.
-
Have the authors considered using additional target–molecule data during adaptation? I wonder whether the limited target-side supervision contributes to the performance gap between T-only and P+T.
6 comments
Response to C5NB. CW 2 and Q2: Pathway ablations and training strategy
1. Pathway ablations and model revision.
We compared the original model with -only, -only, and Shared variants. The -only variant disables the residual pathway; -only suppresses the phenotype-derived prompt on the P-only route; Shared replaces the separate projection heads with one shared projection.
| OOD split | Route | Original Full | -only | -only | Shared |
|---|---|---|---|---|---|
| Chemical | P-only | 0.578 | 0.579 | 0.569 | 0.572 |
| Scaffold | P-only | 0.570 | 0.575 | 0.567 | 0.571 |
| Cell-line | P-only | 0.552 | 0.557 | 0.550 | 0.550 |
| Target | P-only | 0.549 | 0.555 | 0.550 | 0.555 |
| Chemical | P+T | 0.557 | 0.584 | 0.559 | 0.562 |
| Scaffold | P+T | 0.555 | 0.584 | 0.559 | 0.559 |
| Cell-line | P+T | 0.539 | 0.563 | 0.532 | 0.538 |
| Target | P+T | 0.543 | 0.566 | 0.540 | 0.543 |
Entries are PRNet from the original forward-KL configuration. In the -only variant, the provided-target route uses alone because replaces and the residual pathway is disabled.
The -only variant matches or exceeds Full in PRNet across all four OOD splits, while -only performs below -only throughout. Shared is also competitive with Full. These results show little additional PRNet benefit from the original residual pathway and motivated us to make its contribution adjustable through a learned scale . Specifically, we added a global scale
to the residual and conditional LoRA updates. Both conditioning interfaces are retained, while their residual contribution is adjusted during training. The scale is shared across both routes and initialized at .
We have adopted this residual-scaled model with reverse KL and updated the main results and component ablations accordingly in our new manuscript. The KL comparison is reported in our response to R1 Q3.
2. When alignment and staged training help.
We agree that KD-only and joint training remain competitive on L1000. To better clarify the contribution of alignment and staged training, we compare these configurations with our full model on both L1000 and Tahoe under P-only inference.
| Method | L1000 (4-split avg.): PRNet ↑ | L1000: VUN ↑ | Tahoe Cell-line OOD: PRNet ↑ | Tahoe Cell-line OOD: VUN ↑ |
|---|---|---|---|---|
| KD-only | 0.582 | 94.4% | 0.225 | 52.8% |
| Joint | 0.581 | 94.2% | 0.383 | 75.4% |
| PhenTTP (Full) | 0.583 | 95.2% | 0.472 | 93.3% |
KD-only retains distillation but removes phenotype–target alignment, while Joint optimizes the training objectives simultaneously. The L1000 results are macro-averaged across four OOD splits.
On L1000, all three configurations achieve comparable performance, suggesting that the simpler training strategies can be sufficient for this setting. However, the differences become more pronounced on the more challenging Tahoe Cell-line OOD split. Our full model achieves substantially higher phenotype agreement (0.472 vs. 0.225/0.383) and VUN (93.3% vs. 52.8%/75.4%) than KD-only and Joint, respectively.
These results suggest that alignment and staged training become more beneficial under challenging distribution shifts, while their additional gains on L1000 are relatively modest (Tahoe: approximately 60,000 gene features vs. L1000: 978 landmark genes). The detailed results are provided in Table 3 and Appendix Tables 8 and 11 of the revised manuscript.
Response to C5NB. Q3 and Q4
Q3: Implementation on ProtoBind-Diff
We thank the reviewer for pointing out this missing implementation detail.
Shared design. In the full -gated formulation, both backbones use a phenotype encoder and a split projection producing and . P-only supplies to the native target-conditioning interface; P+T supplies , while retaining the same branch. The trainable components are the phenotype encoder, split projection, gated residual modules, conditional LoRA in the last two decoder layers, and a global scalar . The pretrained backbone and native target pathway remain frozen. Initialized to , is shared across layers and routes and scales both residual and LoRA updates, without scaling the native target condition.
DrugGPT adapter. The native interface is the autoregressive protein-prompt slot: supplies four soft prompt tokens in the model's token-embedding space, or an explicit protein prompt is used in P+T mode. The pooled conditions an added gated residual module at the adapter slot within each Transformer block. A query mask restricts its updates to states predicting molecular tokens, preserving the native protein-prompt computation.
ProtoBind-Diff adapter. The native interface is the protein-conditioning memory of the masked-diffusion decoder. Four 1,280-dimensional tokens replace the ESM-2 protein embeddings in P-only mode; P+T uses the native protein embeddings. Both pass through the frozen sequence projection into the existing cross-attention memory, together with the diffusion-noise token. In the full formulation, the -conditioned adapter adds its gated residual and applicable LoRA update after each denoising decoder block, using the same shared rule. Thus, the module roles are shared, while the native conditioning interface and residual insertion points are backbone-specific.
Q4: Target-wise comparison and training coverage
We examined paired test conditions against DrugGPT, with both models receiving the same target. We report the distribution of wins and its relationship with target-annotated training coverage below.
1. Distribution of performance differences.
For PRNet, we average scores within each target and count which model performs better. DrugGPT wins slightly more than half of the targets in Scaffold, Cell-line and Target OOD, while Combined wins slightly more than half in Chemical OOD. Thus, the target-wise win frequencies are close to split evenly.
| OOD split | PRNet target wins: Combined | PRNet target wins: DrugGPT | Vina condition wins: Combined | Vina condition wins: DrugGPT | Boltz-2 condition wins: Combined | Boltz-2 condition wins: DrugGPT |
|---|---|---|---|---|---|---|
| Chemical | 60/116 | 56/116 | 45/100 | 54/100 | 46/100 | 53/100 |
| Scaffold | 55/119 | 64/119 | 47/100 | 52/100 | 43/100 | 56/100 |
| Cell-line | 246/501 | 255/501 | 51/100 | 46/100 | 53/100 | 44/100 |
| Target | 28/60 | 32/60 | 48/100 | 50/100 | 41/100 | 57/100 |
Combined denotes the earlier model's provided-target route. PRNet uses all 2,000 phenotype requests per split. Affinity wins are counted over all 100 requested conditions per split; 99/99/97/98 conditions have jointly finite scores in Chemical/Scaffold/Cell-line/Target OOD, respectively. Lower Vina and higher Boltz-2 indicate a win. Ties and unscored requests award neither model a win and remain in the denominator.
These counts describe how frequently each model performs better. They do not quantify whether a few large target-specific differences dominate the aggregate gap. The affinity entries are condition-level comparisons, rather than target-win counts.
2. Relationship with target-annotated training coverage.
We relate each target's PRNet difference, defined as Combined minus DrugGPT, to its number of distinct target-annotated training molecules using Spearman correlation. The results are () for Chemical, () for Scaffold, and () for Cell-line OOD.
We find no clear monotonic association between target-specific training coverage and the PRNet difference in these splits. For Target OOD, every held-out target has zero target-specific training coverage, so the correlation is undefined.
Response to C5NB. CW 1 and Q1: Phenotype control and the role of P+T
1. Our task is phenotype-driven molecular generation.
Our goal is to generate molecules from a requested cellular response, without requiring a known target. Native DrugGPT requires a target and does not provide a trained phenotype-conditioning interface.
We tested direct phenotype input to unmodified DrugGPT by supplying numeric expression values through its native protein prompt slot, without a target or further training. Under its native context limit, 153–188 of the 978 gene values are retained after reserving generation tokens.
Each method has 2,000 requests per split. Only 59 of 8,000 DrugGPT phenotype-text outputs are valid (0.74), all representing
CN. Thus, its unmodified native interface does not provide effective phenotype-driven generation.2. P+T adds target guidance while retaining phenotype control.
P+T is an additional function built on our phenotype-driven framework: users can specify both the desired cellular response and a target. Its value is that the phenotype continues to influence generation when the target is fixed.
We test this using the original P+T checkpoint. For 2,000 requests per split, we keep the target unchanged and replace the full phenotype input with a frozen dissimilar donor. Both outputs are evaluated against the original requested phenotype in its original cell, dose and time context.
Entries are matched/replaced phenotype input. The matched input gives higher correlations in all twelve evaluator-by-split point-estimate comparisons. All requests are retained, including the original invalid-output penalties. Both diagnostic experiments above use the original forward-KL configuration.
These results show that phenotype conditioning improves fulfillment of the requested response at a fixed target. T-only does not provide this control: its conditional distribution is unchanged across different phenotype requests. Its aggregate phenotype scores can nevertheless be high because target identity is associated with cellular responses.
3. Joint supervision is limited.
The main data limitation is the availability of linked phenotype–target–molecule supervision. In the constructed L1000 dataset, only 2,127 of 17,766 molecules have usable target-sequence annotations. Phenotype–molecule training can use the full pair pool, whereas joint conditioning relies on this much smaller annotated subset.
Sparse joint coverage may leave phenotype-dependent adaptation insufficiently constrained by target supervision. The resulting updates can perturb the pretrained target-conditioned prior, potentially reducing both phenotype agreement and affinity relative to T-only. This is a plausible contributor to the original P+T performance gap.
There is also a trade-off between phenotype agreement and target affinity when adding target guidance. Improving this balance with broader joint supervision remains an important direction for further work. The revised model retains rP with adjustable strength, rather than removing phenotype conditioning from the provided-target route.
4. DrugGPT is a target-only reference.
DrugGPT is included as a reference for target performance, not a baseline for our phenotype-driven task. The direct task baselines are phenotype-conditioned generators. DrugGPT shows what the pretrained target expert achieves when a target is supplied.
The revised model uses controlled residual adaptation and reverse KL. Its P+T affinity results are now competitive with the DrugGPT reference:
These updated results are reported in Appendix Table 9. PhenTTP supports a broader discovery setting: generation from a desired cellular response when the target is unknown, and additional target guidance when one is available. The target-only reference measures how much target performance is retained within this phenotype-driven framework.