acceptodds
Under review as a conference paper at ICLR 2027

PEAC: Joint Generation of Promoter-Enhancer Pairs with Activity Conditioning

Abstract

Designing regulatory DNA with predictable gene-expression activity remains challenging since promoters and enhancers usually interact as pairs, while current approaches always focus on the design of individual regulatory elements. Ignoring promoter–enhancer interactions results in regulatory elements that perform well individually but fail to achieve the desired activity when combined. To address this problem, we propose PEAC, a Promoter-Enhancer Activity-Conditioned co-design framework that achieves joint modeling of promoter-enhancer pairs. Specifically, we first train a student model that is distilled from a promoter teacher and an enhancer teacher simultaneously, and then fine-tune the student model with the activity constraint to achieve the conditional generation of the target activity. Furthermore, we also incorporate pair-level fusion and latent distribution alignment to make the generated promoter-enhancer pairs better matched. Extensive experiments show that PEAC achieves excellent performance in generating activity conditioned promoter-enhancer pairs with sequence realism. In oracle-based evaluation, PEAC demonstrated the ability to generate pairs across low-, medium-, and high-activity regimes, with target regime fractions of 79.1%, 91.7% and 67.8%, compared with 31.8%, 68.6%, and 11.6% for baseline. This improvement is accompanied by realistic GC content, high motif enrichment, and low biological distance compared to reference. Together, these results establish PEAC as a native promoter-enhancer generator that makes pair-level activity the explicit design target and provides an oracle-evaluated framework for activity-conditioned regulatory sequence co-design.

Primary area:
applications to physical sciences (physics, chemistry, biology, etc.)
Keywords:
Regulatory sequence design, promoter-enhancer pairs, activity-conditioned DNA generation, DNA language models, latent distribution alignment

Reviews

Among:
Shape:
1234this paper, Bottom 27%
Figure 1. Mean review rating of the reviewed papers at ICLR 2027. This paper averages 2.00 over 3 reviews, above 27% of the others.
Reviewer_GZwYOpenReview ↗
Rating
2
Confidence
2
Informativeness
2
Presentation
2
Soundness
2
Read review
Paper summary

This paper proposes PEAC, a framework for activity-conditioned joint generation of promoter–enhancer pairs. PEAC consists of three stages: regulatory prior distillation from promoter and enhancer teacher models into a shared autoregressive Transformer, activity-conditioned initialization using STARR-seq-derived element-level supervision, and pair fusion with bidirectional cross-attention and conditional maximum mean discrepancy (MMD) alignment using FANTOM5 CAGE-derived pair-level data. The authors evaluate PEAC on K562-related datasets using single-element activity predictors, a CAGE-expression oracle, sequence realism and diversity metrics, ablation studies, constructed-pair baselines, and multiple training and generation seeds. Compared with a direct-concatenation baseline, PEAC achieves higher target-bin hit rates under the primary pair-activity oracle for low-, medium-, and high-activity conditions (79.1%, 91.7%, and 67.8%, respectively, versus 31.8%, 68.6%, and 11.6%). The authors also report competitive sequence statistics and closer matching to the reference activity distribution than several constructed-pair baselines. However, independent-oracle evaluation reveals substantially weaker low-activity controllability, and an external EPInformer diagnostic shows little difference between matched and shuffled pairs, although the same limitation is observed for reference pairs. Overall, the results provide encouraging evidence for oracle-based activity-conditioned pair generation, but whether PEAC learns meaningful promoter–enhancer functional compatibility or produces biologically functional sequences remains insufficiently established.

Critical strengths
  1. The paper addresses a meaningful and relatively underexplored problem. Extending regulatory DNA generation from individual promoters or enhancers to activity-conditioned promoter–enhancer co-generation is a relevant and potentially valuable research direction, given that regulatory activity depends on the combination of these elements. The problem formulation is clear, and the autoregressive factorization provides a natural way to generate an enhancer conditioned on both the target activity and the generated promoter. The work therefore offers a useful formulation for studying joint regulatory sequence generation.

  2. The evaluation covers multiple aspects of generation quality and includes useful robustness checks. Beyond evaluating activity controllability with the primary CAGE-expression oracle, the paper reports sequence-level realism and diversity diagnostics, component ablations, constructed-pair baselines, independent-oracle evaluations, and multi-seed experiments. These experiments provide a broader assessment than relying on a single activity metric or a single model configuration. In particular, the independent-oracle analysis and constructed-pair controls help characterize the limitations of the reported activity-control results and provide a useful starting point for evaluating future methods for promoter–enhancer co-generation.

Critical weaknesses
  1. The central claim of pair-specific functional modeling is not sufficiently established. The main evidence for activity controllability relies on a CAGE-expression oracle trained on the same pair-level data distribution used for model development. Although the authors evaluate the generated samples using three additional oracles trained with leakage-controlled splits, the low-activity target-bin hit rate drops from 79.1% under the primary oracle to 35.1% under majority voting across the independent oracles. Furthermore, the EPInformer diagnostic shows nearly identical predicted activity and pair-attention scores for matched and shuffled PEAC pairs. While the similar results for reference pairs suggest that this diagnostic may have limited sensitivity to pair-specific compatibility, it does not provide positive evidence that PEAC captures such compatibility either. Taken together, the results support some degree of oracle-predicted activity control, but do not sufficiently establish that the model learns meaningful promoter–enhancer functional compatibility beyond marginal sequence properties and condition-dependent generation. This limitation directly affects the central motivation and biological interpretation of the proposed framework.

  2. The contribution of pair fusion is difficult to isolate, and potential shortcuts in the training objectives remain insufficiently controlled. The pair-fusion encoder and auxiliary prediction heads receive explicit target-condition information, while generated and reference representations for the MMD objective are computed using the same jointly trained encoder. Consequently, the activity-prediction heads may partially exploit the supplied condition, and the MMD objective may align representations through condition information rather than sequence-derived functional features. The authors acknowledge this potential shortcut in Appendix A.7, but the reported ablations and external-oracle evaluations do not directly rule it out. Separately, although the direct-concatenation baseline uses the same pair-level training data and activity labels, its modeling and training pipeline differs from that of PEAC, making it difficult to attribute the observed improvements specifically to cross-attention pair fusion. A condition-blind evaluation and a more closely matched baseline that differs primarily in the pair-fusion mechanism would be needed to establish the specific contributions of these components. Without such evidence, the extent to which pair fusion and latent distribution alignment enforce sequence-dependent functional compatibility remains unclear, weakening the methodological claims supported by the experiments.

Questions for authors
  1. Robustness of activity conditioning. The low-activity target-bin hit rate decreases from 79.1% under the primary CAGE-expression oracle to 35.1% under majority voting across the three leakage-controlled oracles (Appendix Tables 7–8). The authors attribute this discrepancy partly to differences in oracle calibration near the low/medium decision boundary. Can the authors clarify how much of the discrepancy can be explained by this boundary effect, and whether the low-activity conditioning result remains robust under continuous-score comparisons or a calibration-independent evaluation? More broadly, how should the primary-oracle results be interpreted in light of the substantially weaker agreement across independent oracles?

  2. Evidence for pair-specific functional compatibility. PEAC is motivated by the hypothesis that promoter–enhancer co-design benefits from modeling functional interactions between the two elements. However, EPInformer assigns nearly identical predicted activity and pair-attention scores to matched and shuffled PEAC pairs, with similar results for reference pairs (Appendix Table 10). What evidence supports the claim that PEAC learns meaningful promoter–enhancer compatibility, rather than primarily learning marginal sequence distributions and target activity labels? In particular, can the authors clarify what the current experiments establish about pair-specific functional compatibility, as distinct from the generation of plausible individual elements and the control of oracle-predicted pair activity?

  3. Potential condition-information shortcuts. The pair-fusion encoder and auxiliary prediction heads receive explicit target-condition information, and the same jointly trained encoder is used to compute generated and reference representations in the MMD objective. Consequently, the activity-prediction heads may exploit the supplied condition, while the MMD loss may align representations partly through condition information rather than sequence-derived functional features. What evidence establishes that these objectives impose meaningful constraints on the generated sequences themselves? For example, have the authors evaluated condition-blind or sequence-only representations, or tested whether the predicted activity and latent representations change appropriately when promoter–enhancer sequences are shuffled while the target condition is held fixed? If such controls are unavailable, how should the claimed contributions of pair fusion and latent distribution alignment be interpreted?

  4. Attribution of improvements to pair fusion. The direct-concatenation baseline differs from PEAC in its training and modeling pipeline, making it difficult to isolate the contribution of explicit pair-level modeling. Although the ablation study reports degradation when pair fusion is removed, can the authors clarify exactly which architectural and training components are changed in this ablation and which are held fixed? In particular, how much of the improvement can be attributed specifically to cross-attention between promoter and enhancer representations, rather than differences in supervision, optimization, or model capacity? A parameter-matched autoregressive baseline trained on the same pair data and activity labels, but without cross-attention pair fusion, would help establish the incremental value of the proposed mechanism.

3 comments

We thank the reviewer for raising these questions. The additional analyses support a narrower conclusion: PEAC jointly generates promoter and enhancer sequences with plausible element-level statistics and can shift sequence-only predictions of pair-reporter activity, with the clearest evidence in the high-activity regime. We have revised the manuscript for clarification.

Weakness 1: Thank you for highlighting this distinction. PEAC aims to jointly generate promoter–enhancer sequences with a requested pair-level activity; it does not assume or claim partner-specific compatibility or interactions beyond multiplicative effects. Accordingly, matched pairs outperforming shuffled pairs is not a necessary criterion for our generation objective. We have clarified this scope in the manuscript and revised wording that could imply biological compatibility. We also explicitly discuss evaluator dependence: low-condition accuracy decreases from 79.1% to 35.1%, whereas the high-condition advantage remains substantial under majority voting across the additional oracles (78.2% versus 12.0% for direct concatenation). The revised discussion treats EPInformer as an inconclusive diagnostic for partner specificity and acknowledges that the additional oracles cannot eliminate shared dataset bias.

Weakness 2: We agree that the existing comparisons do not fully isolate the contribution of pair fusion or exclude condition-based shortcuts. We have revised the methodological interpretation accordingly: cross-attention enables information exchange between element representations, and MMD provides an auxiliary distributional constraint; neither is presented as proof of sequence-dependent functional compatibility. We have also clarified that the direct-concatenation comparison establishes the performance difference between the evaluated pipelines. Condition-blind auxiliary objectives and a baseline matched in initialization, losses, and training budget remain necessary controls for stronger component-level conclusions, and this limitation is now stated explicitly.

Question 1: We found that the reported 79.1% low-bin hit rate came from a stale score summary. On the identical 1,000 saved low-conditioned pairs used for disjoint-oracle rescoring, the primary oracle gives 48.7% and majority voting gives 35.1%. Comparing inconsistent result artifacts accounted for 30.4 percentage points of the original 44.0-point gap. The remaining same-sample difference is 13.6 points. The primary oracle predicts Bergman K562 ExP STARR-seq log₂(RNA/DNA) pair-reporter activity. CAGE-derived supervision is used during initialization.

The generator and evaluators use different condition edges, and the primary and disjoint oracles also differ in binning. Predictions close to the low/medium boundary can therefore change bins. The current analysis does not estimate what fraction of the remaining 13.6 points is attributable to the boundary alone.

We fit a monotone calibration map separately on each disjoint oracle’s validation predictions and froze it before scoring the generated pairs. The calibrated PEAC means for low, medium, and high targets were −4.964, −4.945, and −1.952. Continuous target MAEs were 1.720, 1.060, and 0.861 for PEAC and 1.637, 0.984, and 2.150 for direct concatenation. The low and medium PEAC distributions nearly coincide, and PEAC improves continuous target error for the high condition. A threshold-free comparison on raw oracle scores found that a randomly drawn high-conditioned pair scored above a randomly drawn low-conditioned pair with probability 0.987–0.989 across the three disjoint evaluators.

The reconciled primary-oracle hit rates describe performance under that evaluator’s decision rule. The strongest cross-oracle result is the upward shift for high-conditioned samples. A robust low-versus-medium distinction is unsupported by the continuous analysis.

Question 2: Our design goal is for the two generated elements to yield a combined reporter activity near the requested target. It does not require the original partners to outperform alternative partners. We recognize that phrases such as “better matched” can imply a stronger goal and will clarify the wording.

In Table 10, EPInformer predicted essentially the same mean activity and pair attention after swapping enhancers in both generated and reference pairs. The reference control shows that weak sensitivity to shuffling is shared by both groups under EPInformer. The test leaves open whether alternative partners have similar effects in this setting or whether the evaluator has limited sensitivity to the swaps. We assess activity target attainment separately with sequence-only reporter oracles.

We also tested partner preference with a frozen evaluator that can detect some non-additive effects in measured Bergman data. On 54,545 held-out combinations it achieved RMSE 1.1508, compared with 1.2191 for an additive promoter-plus-enhancer model on the log-activity scale. Its predicted and measured difference-in-differences correlated at Pearson r = 0.373. On generated pairs, we fixed the promoter and swapped enhancers within the same requested bin and nearby activity quantiles. The full model’s mean matched-minus-crossed prediction was 0.00193 (95% training-seed-clustered bootstrap interval −0.00220 to 0.00527), with a matched win rate of 50.12%. The full-minus-no-fusion contrast was 0.00140 (95% interval −0.02366 to 0.02975).

Question 3: We tested three independently trained checkpoints each for full PEAC, no pair fusion, and no MMD. We permuted 10-bp blocks within one element at a time while holding the scalar target, activity bin, and within-bin quantile fixed. In the converse control we held both DNA sequences fixed and changed only the condition. These within-element perturbations probe sequence sensitivity. The activity-matched enhancer exchanges in response 2 address partner preference.

For full PEAC, the mean absolute change in the internal activity prediction was 0.0025 activity units after sequence perturbation and 4.670 after changing only the condition. The corresponding latent RMS changes were 0.0099 and 1.818. The bin head remained 100% accurate after sequence perturbation. These results indicate that the internal heads and the MMD representation depend strongly on the supplied condition under this audit.

We also evaluated the same sequences with a separately trained, frozen Bergman predictor that accepts only DNA. Its full-model scores had mean Spearman correlation 0.621 ± 0.187 with the requested activity across training seeds. Sequence perturbation changed its score by 0.396 ± 0.024 activity units and increased target MAE by 0.022 ± 0.008. Thus, generated sequences carry activity information detectable by a sequence-only predictor. Full PEAC did not consistently outperform the no-fusion or no-MMD variants on these frozen-evaluator outcomes. We interpret both objectives as condition-aware auxiliary regularizers, with their incremental sequence-level contribution still unresolved.

Question 4: The closest isolation test is the matched no-pair-fusion variant. It retains the same Bergman pair rows and labels, condition embeddings, language-model and auxiliary losses, and final-stage optimization schedule of six epochs, batch size 8, learning rate 5 × 10⁻⁵, and up to 7,200 updates. Both variants have 18,095,018 nominal training-time parameters and the same 7,408,912-parameter causal rollout architecture. The ablation disables bidirectional promoter–enhancer state exchange and the pair token’s attention to the sequence states. Its disabled weights remain in the parameter count. Both variants retain promoter-first autoregression, so the enhancer can depend on the promoter during generation. The ablation tests the disabled fusion paths together, and the changes cannot be attributed to either path individually.

Across three independent training seeds, after averaging the three generation seeds within each checkpoint, the paired full-minus-no-fusion changes in low, medium, and high hit rates were −13.1 ± 19.9, +6.9 ± 7.3, and +6.3 ± 7.1 percentage points (mean ± SD). In these runs, fusion raised the mean medium and high rates and lowered the mean low rate. The seed variability and partner-swap results limit stronger attribution.

The direct-concatenation Transformer uses the same pair table and update count and has 7,408,656 rollout parameters. It uses a different training pipeline, batch size 16, and learning rate 3 × 10⁻⁴. It provides a conventional autoregressive benchmark. A conventional baseline matched on the training and example budgets, optimization, supervision, and seeds would be needed to isolate the incremental effect against that model family.

Reviewer_1sT3OpenReview ↗
Rating
2
Confidence
2
Informativeness
2
Presentation
2
Soundness
1
Read review
Paper summary

This paper proposes a method to jointly generate promotor and enhancer conditioned on activity. The model went through distillation, single element activity conditioned generation and joint generation. The model is evaluated using an oracle model trained on the same dataset used for the joint training.

Critical strengths

The paper did activity conditioned joint generation of promotor and enhancer pairs, and shows a strong engineering effect at training the model in stages and doing different ablation studies.

Critical weaknesses

Earlier work (Bergman et al. 2022) has shown that the pair activity could be largely explained by a multiplicative combination of promoter and enhancer activities, if so the conditional generation could largely be achieved by activity pairing. The paper didn't provide enough evidence that PEAC captures interactions beyond the multiplicative effect. The main evaluation relies on an oracle model trained on the same underlying CAGE-derived dataset, which could capture the same dataset bias. EPInformer also didn't find matched PEAC pairs more active than shuffled pairs, and the test of EPInformer on real reference pairs was not stratified by activity level. I think you can do an additional test on Bergman's dataset through GEO (GSE184426).

Questions for authors

See the previous section.

Reviewer_FLaxOpenReview ↗
Rating
2
Confidence
4
Informativeness
2
Presentation
2
Soundness
2
Read review
Paper summary

This paper introduces PEAC, a framework for activity-conditioned joint generation of promoter–enhancer pairs. Unlike existing regulatory DNA generation approaches that primarily focus on individual promoters or enhancers, PEAC aims to generate both regulatory elements jointly while controlling their predicted pair-level activity. The framework consists of three stages: regulatory prior distillation from pretrained promoter and enhancer teachers, activity-conditioned initialization using STARR-seq-derived supervision, and pair-level fine-tuning using FANTOM5 CAGE-derived expression labels. The final stage incorporates bidirectional cross-attention for promoter-enhancer representation fusion and conditional maximum mean discrepancy (MMD) alignment to encourage generated pairs to match reference distributions within specified activity regimes.

Critical strengths
  • The first critical strength is that the paper addresses an important and relatively underexplored problem in regulatory DNA generation: the joint design of promoters and enhancers with controllable pair-level activity. Most existing generative approaches optimize individual regulatory elements, even though transcriptional activity depends on interactions between promoters, enhancers, and their cellular context.

  • The second critical strength is the relatively comprehensive empirical evaluation of the proposed framework. The paper examines functional activity, conditional controllability, sequence realism, and diversity, rather than relying exclusively on predicted activity improvements.

Critical weaknesses
  • The first major weakness is that the paper does not sufficiently establish that the proposed framework learns meaningful promoter-enhancer compatibility beyond activity-conditioned sequence generation. Although PEAC incorporates cross-attention and latent distribution alignment, these architectural components do not necessarily demonstrate that the generated promoter and enhancer form a functionally coordinated pair. In particular, Appendix Table 10 shows that replacing the generated enhancer with an enhancer from another generated pair produces almost no change in EPInformer-predicted pair activity or attention. The matched-versus-shuffled activity difference is approximately −0.000016, and the fraction of matched pairs outperforming shuffled pairs is only 0.501. Although reference pairs exhibit similarly weak separation under this diagnostic, the result does not provide positive evidence that PEAC captures meaningful pair-specific interactions. Furthermore, the model generates promoters before enhancers using an autoregressive factorization, while the cross-attention fusion module primarily operates on learned representations during training. The manuscript does not clearly establish whether the improvements arise from genuine promoter-enhancer interaction modeling or from activity-conditioned generation and stronger sequence priors. This distinction is consequential because joint regulatory compatibility is the central motivation and claimed contribution of the paper.

  • The second major weakness is that the biological validity and generalization of the reported improvements remain insufficiently demonstrated. The main evaluation relies on a self-trained CAGE-expression oracle, while the underlying supervision consists of processed CAGE-derived expression scores assigned to candidate promoter-enhancer pairs rather than direct measurements of the activity of the generated pairs. Although the primary oracle achieves a held-out Spearman correlation of 0.868, its performance decreases substantially under promoter-disjoint and enhancer-disjoint evaluations, where Spearman correlations fall to 0.683 and 0.696, respectively. More importantly, Appendix Table 8 shows that the reported controllability improvements are not consistently reproduced by independently trained oracles. For example, the low-activity target-bin fraction decreases from 0.791 under the primary oracle to 0.351 under majority voting across disjoint oracles, while the direct concatenation baseline outperforms PEAC for medium-activity conditioning under the same independent-oracle evaluation (0.851 versus 0.775). These results suggest that some of the reported gains may depend on the calibration or inductive biases of the evaluation predictor. The external EPInformer analysis provides a useful complementary diagnostic but does not establish experimentally measured promoter–enhancer function. Furthermore, the experiments are restricted primarily to K562 and fixed-length regulatory sequences, without validation across additional cell types, genomic contexts, or experimental assays. Consequently, the current results support computational activity-regime control more strongly than robust biological generalization or functional regulatory sequence co-design.

Questions for authors
  • Could the authors clarify what evidence demonstrates that PEAC learns genuine promoter–enhancer compatibility rather than primarily activity-conditioned sequence generation? The proposed pair-fusion module introduces bidirectional cross-attention and latent alignment, but Appendix Table 10 reports almost identical EPInformer activity and attention scores for matched and shuffled promoter–enhancer pairs. This raises the question of whether the generated pairs exhibit meaningful interaction-specific behavior. A more direct comparison between matched and shuffled pairs using a validated pair-sensitive predictor, together with an analysis of whether the predicted effect of an enhancer depends on its promoter partner, would materially affect my assessment. It would also be useful to compare PEAC against a simpler autoregressive conditional generator without explicit pair fusion under matched training conditions. Such experiments could help isolate whether the proposed pair-level architecture captures interactions beyond the promoter-first autoregressive factorization.

  • Could the authors clarify how the pair-fusion and latent MMD objectives ensure that activity supervision depends on the generated DNA sequences rather than the explicitly provided activity conditions? As acknowledged in Appendix A.7, the auxiliary pair encoder receives the target activity condition, activity bin, and activity quantile, which may allow the activity-prediction and classification heads to recover the conditioning target without learning the corresponding sequence-dependent regulatory function. Similarly, because both generated and reference latent representations are produced by a jointly updated condition-aware encoder, minimizing MMD does not necessarily imply that the generated sequences become more biologically realistic or functionally compatible. A controlled experiment using a condition-blind or frozen pair evaluator, together with promoter/enhancer perturbation tests under fixed activity conditions, would help determine whether these objectives provide genuine sequence-level supervision. This issue is particularly important because the pair-fusion and distribution-alignment components constitute the main methodological contributions beyond conventional conditional autoregressive generation.

  • Could the authors provide a more controlled evaluation of activity-conditioned generation using independent predictors and explain the substantial differences between the primary-oracle and leakage-controlled results? In Table 2, PEAC achieves strong target-bin fractions across all three activity regimes. However, Appendix Table 8 shows that the low-condition fraction decreases substantially under independent oracles, and the direct concatenation baseline achieves a higher medium-condition fraction under majority voting. Although the manuscript discusses calibration differences and activity-bin boundaries, it remains unclear how much of the reported improvement reflects robust changes in generated regulatory activity rather than predictor-specific decision boundaries. Reporting continuous activity distributions, calibration-adjusted target errors, and confidence intervals across independently trained oracles would help clarify this issue. In addition, the authors should explain how candidate-pair labels are constructed from the processed CAGE data and whether the reported scores can be interpreted as pair-specific regulatory activity rather than promoter-associated expression proxies. These clarifications would materially affect the interpretation of the central experimental results.

  • Could the authors clarify the contribution of each training stage and explain the inconsistencies between the single-run and multi-seed ablation results? Table 3 suggests that removing activity-conditioned initialization, pair fusion, or latent MMD generally degrades activity-related metrics and sequence realism. However, Appendix Tables 16 and 18 reveal a more complicated picture. For example, removing pair fusion improves the mean low-condition hit rate from 0.571 to 0.702, while removing latent MMD improves the mean high-condition hit rate from 0.640 to 0.670. In addition, the supervision-source stress test in Appendix Table 19 shows that removing distillation can improve several sequence-realism metrics. These observations suggest that the proposed components may introduce trade-offs rather than consistently improving controllability. It would be valuable to report matched-capacity and matched-training-budget comparisons, including a direct conditional generator trained only on CAGE-derived pairs, and to explain which components are necessary for pair-specific modeling rather than primarily preserving sequence realism. A clearer analysis of these trade-offs would strengthen the justification for the proposed three-stage architecture.

  • Could the authors provide stronger evidence that the generated promoter–enhancer pairs generalize to biologically meaningful regulatory settings beyond the current K562 computational benchmark? The current experiments use fixed 250-bp promoter and 230-bp enhancer windows and do not explicitly model genomic distance, chromatin accessibility, three-dimensional contacts, or broader cell-type-specific regulatory context. These factors can substantially influence promoter-enhancer interactions and may limit the biological interpretation of pair-level activity predictions. Although experimental validation may be beyond the scope of the current submission, evaluation on an independent experimentally measured promoter–enhancer compatibility dataset, or testing whether generated enhancers exhibit promoter-dependent effects under a separately validated model, would provide substantially stronger evidence for the claimed joint-design capability. The authors should also clarify whether PEAC is intended primarily as a framework for matching computationally defined activity distributions or as a method for generating biologically functional and compatible regulatory element pairs, since these objectives require different levels of validation.

From OpenReview, CC BY 4.0. Authors as signed there; shown unedited.

Related papers

Loading the map…

Discussion (0)

Sign in to comment.