pith. sign in

arxiv: 2606.21109 · v1 · pith:B7XKQQOJnew · submitted 2026-06-19 · 💻 cs.LG · cs.AI

Chem2Gen-Bench: Benchmarking Chemical-to-Genetic Translation in Perturbation Response Space

Yuxiang Lin , Ying Chen This is my paper

Pith reviewed 2026-06-26 14:41 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords chemical perturbationgenetic perturbationbenchmarkperturbation responseretrieval taskembedding comparisonK562virtual cell
0
0 comments X

The pith

Chemical-to-genetic perturbation translation shows measurable but heterogeneous fidelity in matched cell contexts

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces Chem2Gen-Bench to test translation between chemical and genetic perturbations by organizing large numbers of response profiles into cell-target matched groups. It measures alignment through pairwise similarity, retrieval tasks, protocol associations, and comparisons of feature spaces against foundation-model embeddings. A sympathetic reader cares because virtual-cell models increasingly need to treat chemical and genetic inputs as interchangeable for predicting cellular responses in drug and gene studies. The central results show that fidelity exists yet varies across contexts, background adjustment strengthens some similarity-retrieval links while lowering mean success in paired tests, and the tested embeddings fail to improve consistently over gene-delta baselines in the K562 audit.

Core claim

Across matched contexts, translation fidelity is measurable but heterogeneous; background adjustment increases the association between pairwise similarity and retrieval success, while paired tests show lower mean retrieval success after adjustment under the evaluated settings. In a target-matched K562 audit, the evaluated foundation-model embeddings did not consistently improve over gene-delta baselines.

What carries the argument

Chem2Gen-Bench, a collection of 260,084 chemical and 1,099,045 genetic perturbation profiles grouped into cell-target contexts for direct pairwise alignment, retrieval, and embedding tests.

If this is right

  • Translation fidelity varies by cell-target context and must be audited per context rather than assumed uniform.
  • Background adjustment strengthens the link between similarity and retrieval success but reduces average retrieval rates in paired settings.
  • Foundation-model embeddings require additional matched evidence before they can be expected to outperform simple gene-delta baselines.
  • The benchmark supplies an auditable way to determine when perturbations align around shared targets.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Reducing unmeasured protocol differences across datasets could narrow the observed heterogeneity in translation success.
  • Gene-delta baselines may remain competitive for many practical predictions until larger matched audits are performed.
  • Extending the same matching approach to additional cell types could identify target classes that support more reliable chemical-genetic interchange.

Load-bearing premise

Chemical and genetic perturbation profiles can be meaningfully organized into comparable cell-target contexts that support direct pairwise alignment, retrieval, and embedding comparisons without major unaccounted batch or protocol effects.

What would settle it

A new audit across additional matched cell lines in which background-adjusted similarity scores show no rise in correlation with retrieval success rates, or in which foundation-model embeddings consistently outperform gene-delta baselines.

Figures

Figures reproduced from arXiv: 2606.21109 by Ying Chen, Yuxiang Lin.

Figure 1
Figure 1. Figure 1: Chem2Gen-Bench construction and benchmark design. (A) Workflow from perturbation profiles to representation construction, cell-target contexts, benchmark levels, pairwise alignment, retrieval, robustness classes, and representation audit. (B) Resource composition, including 260,084 chemical perturbation profiles and 1,099,045 genetic perturbation profiles. (C) Pairwise alignment across matched contexts. (D… view at source ↗
Figure 2
Figure 2. Figure 2: Representation contrasts, protocol associations, and embedding audit. (A) Paired representation contrasts for Standard versus Systema-style views and gene versus pathway feature spaces. (B) Descriptive class trajectories across benchmark levels. (C) Dose and time associations with retrieval behavior in selected contexts. (D) Robust-high target enrichment. (E) K562 foundation￾model audit leaderboard by targ… view at source ↗
read the original abstract

Virtual-cell and perturbation models are increasingly used to predict cellular responses for biomedical discovery, but chemical and genetic perturbations are not automatically interchangeable. Existing evaluations often study chemical response prediction or genetic perturbation prediction separately, leaving target-matched chemical-to-genetic translation under-tested. We introduce Chem2Gen-Bench, a benchmark comprising 260,084 chemical and 1,099,045 genetic perturbation profiles organized into cell-target contexts, and evaluate pairwise alignment, retrieval, protocol covariate associations, feature spaces, and foundation-model embeddings. Across matched contexts, translation fidelity is measurable but heterogeneous; background adjustment increases the association between pairwise similarity and retrieval success, while paired tests show lower mean retrieval success after adjustment under the evaluated settings. In a target-matched K562 audit, the evaluated foundation-model embeddings did not consistently improve over gene-delta baselines. Chem2Gen-Bench provides an auditable framework for testing when chemical and genetic perturbations align around shared targets and when representation gains are supported by matched perturbation evidence.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The paper introduces Chem2Gen-Bench, a benchmark with 260,084 chemical and 1,099,045 genetic perturbation profiles organized into cell-target contexts. It evaluates pairwise alignment, retrieval, protocol covariate associations, feature spaces, and foundation-model embeddings. Key claims are that translation fidelity is measurable but heterogeneous; background adjustment increases the association between pairwise similarity and retrieval success while lowering mean retrieval success in paired tests; and in a target-matched K562 audit, evaluated foundation-model embeddings did not consistently improve over gene-delta baselines. The work positions the benchmark as an auditable framework for testing when chemical and genetic perturbations align around shared targets.

Significance. If the matching of profiles into comparable contexts is valid, the benchmark addresses a gap in separate evaluations of chemical vs. genetic perturbation prediction by providing target-matched translation tests at scale. The explicit provision of an auditable framework and large organized dataset counts as a strength for reproducibility. The findings on heterogeneous fidelity and adjustment effects could inform virtual-cell model development if technical confounds are ruled out.

major comments (2)
  1. [Abstract (data organization and evaluation)] The central claims on measurable but heterogeneous translation fidelity, background adjustment effects, and foundation-model performance rest on the assumption that the 260k chemical + 1M genetic profiles can be partitioned into comparable cell-target contexts (shared target gene + cell line) without residual technical variation from protocols, platforms, or batches. The abstract states that protocol covariate associations were evaluated, but it is unclear from the description whether these were corrected or only correlated; if only correlated, the reported heterogeneity and retrieval outcomes could be artifacts. This assumption is load-bearing for all empirical results.
  2. [Abstract (K562 audit)] In the target-matched K562 audit, the conclusion that foundation-model embeddings did not consistently improve over gene-delta baselines is directly sensitive to any platform mismatch between chemical and genetic profiles. Without explicit validation that matching criteria eliminate systematic offsets, the comparison between embeddings and baselines cannot be interpreted as evidence about representation gains.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive comments on Chem2Gen-Bench. We address each major comment below with clarifications on our methods and indicate where revisions will be made to improve transparency.

read point-by-point responses

  1. Referee: [Abstract (data organization and evaluation)] The central claims on measurable but heterogeneous translation fidelity, background adjustment effects, and foundation-model performance rest on the assumption that the 260k chemical + 1M genetic profiles can be partitioned into comparable cell-target contexts (shared target gene + cell line) without residual technical variation from protocols, platforms, or batches. The abstract states that protocol covariate associations were evaluated, but it is unclear from the description whether these were corrected or only correlated; if only correlated, the reported heterogeneity and retrieval outcomes could be artifacts. This assumption is load-bearing for all empirical results.

    Authors: We agree that the distinction between correlation and correction is important for interpretation. The manuscript evaluates associations between protocol covariates and the similarity/retrieval metrics (as stated in the abstract) but does not apply any corrections to the profiles. This choice was deliberate to create an auditable benchmark that surfaces the effects of technical variation rather than removing them. We will revise the abstract, methods, and results sections to explicitly state that associations were quantified without correction, and we will add further discussion of how users can apply their own adjustments within the framework. revision: yes

  2. Referee: [Abstract (K562 audit)] In the target-matched K562 audit, the conclusion that foundation-model embeddings did not consistently improve over gene-delta baselines is directly sensitive to any platform mismatch between chemical and genetic profiles. Without explicit validation that matching criteria eliminate systematic offsets, the comparison between embeddings and baselines cannot be interpreted as evidence about representation gains.

    Authors: We acknowledge that the K562 results are conditional on the target-matched context definition and that no additional post-matching validation for platform offsets was performed beyond the shared cell line and target criteria. The benchmark is presented as an evaluation under these matching rules, with the explicit goal of enabling further audits. We will expand the discussion section to note the potential for residual platform effects and the limitations this imposes on claims about representation gains. revision: partial

Circularity Check

0 steps flagged

Empirical benchmark evaluation with no derivation chain

full rationale

This is a data-organization and retrieval benchmark paper. It partitions existing chemical/genetic profiles into cell-target contexts, computes pairwise similarities and retrieval success, tests protocol covariates, and compares foundation-model embeddings against gene-delta baselines. No equations, first-principles derivations, or predictions are claimed; all reported quantities (fidelity, adjustment effects, K562 audit) are direct empirical measurements on the assembled dataset. No self-citation load-bearing steps, fitted-input-as-prediction, or ansatz smuggling appear. The work is self-contained against external benchmarks and receives the default low-circularity finding.

Axiom & Free-Parameter Ledger

1 free parameters · 0 axioms · 0 invented entities

Abstract-only review means full details on parameters and assumptions are unavailable. The benchmark construction implies unstated choices in data filtering, context matching, and metric definitions that function as free parameters.

free parameters (1)
  • background adjustment protocol
    The abstract evaluates results with and without background adjustment, indicating this is a modeling choice that affects reported associations and retrieval scores.

pith-pipeline@v0.9.1-grok · 5700 in / 1239 out tokens · 34736 ms · 2026-06-26T14:41:47.210012+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

83 extracted references · 74 canonical work pages

  1. [1]

    2024 , pmid =

    Nature Methods , author =. 2024 , pmid =. doi:10.1038/s41592-023-02144-y , abstract =

    work page doi:10.1038/s41592-023-02144-y 2024

  2. [2]

    Nature Methods , author =

    Three million images and morphological profiles of cells treated with matched chemical and genetic perturbations , issn =. Nature Methods , author =. 2024 , pmid =. doi:10.1038/s41592-024-02241-6 , abstract =

    work page doi:10.1038/s41592-024-02241-6 2024

  3. [3]

    Nature Methods , author =

    Causal identification of single-cell experimental perturbation effects with. Nature Methods , author =. 2023 , pmid =. doi:10.1038/s41592-023-02040-5 , abstract =

    work page doi:10.1038/s41592-023-02040-5 2023

  4. [4]

    Nature Medicine , author =

    A foundation model for clinician-centered drug repurposing , issn =. Nature Medicine , author =. 2024 , pmid =. doi:10.1038/s41591-024-03233-x , abstract =

    work page doi:10.1038/s41591-024-03233-x 2024

  5. [5]

    Nature Biotechnology , author =

    Predicting transcriptional outcomes of novel multigene perturbations with. Nature Biotechnology , author =. 2024 , pmid =. doi:10.1038/s41587-023-01905-6 , abstract =

    work page doi:10.1038/s41587-023-01905-6 2024

  6. [6]

    Nature Communications , author =

    Predicting transcriptional responses to novel chemical perturbations using deep generative model for drug discovery , volume =. Nature Communications , author =. 2024 , pmid =. doi:10.1038/s41467-024-53457-1 , abstract =

    work page doi:10.1038/s41467-024-53457-1 2024

  7. [7]

    URL https://doi.org/10

    Learning single-cell perturbation responses using neural optimal transport , volume =. Nature Methods , author =. 2023 , pmid =. doi:10.1038/s41592-023-01969-x , abstract =

    work page doi:10.1038/s41592-023-01969-x 2023

  8. [8]

    2024 , pmid =

    Genome Biology , author =. 2024 , pmid =. doi:10.1186/s13059-024-03436-y , abstract =

    work page doi:10.1186/s13059-024-03436-y 2024

  9. [9]

    Benchmarking

    Li, Chen and Gao, Haoxiang and She, Yuli and Bian, Haiyang and Chen, Qing and Liu, Kai and Wei, Lei and Zhang, Xuegong , month = jan, year =. Benchmarking. doi:10.1101/2024.12.20.629581 , abstract =

    work page doi:10.1101/2024.12.20.629581 2024

  10. [10]

    Nature Communications , author =

    Predicting cell morphological responses to perturbations using generative modeling , volume =. Nature Communications , author =. 2025 , pmid =. doi:10.1038/s41467-024-55707-8 , abstract =

    work page doi:10.1038/s41467-024-55707-8 2025

  11. [11]

    Nature , year=

    Mapping cells through time and space with moscot , copyright =. Nature , author =. 2025 , note =. doi:10.1038/s41586-024-08453-2 , abstract =

    work page doi:10.1038/s41586-024-08453-2 2025

  12. [12]

    Cell , author =

    Multiplexed single-cell characterization of alternative polyadenylation regulators , volume =. Cell , author =. 2024 , pmid =. doi:10.1016/j.cell.2024.06.005 , abstract =

    work page doi:10.1016/j.cell.2024.06.005 2024

  13. [13]

    doi:10.1101/2025.02.05.636714 , abstract =

    Wang, Chloe and Cui, Haotian and Zhang, Andrew and Xie, Ronald and Goodarzi, Hani and Wang, Bo , month = feb, year =. doi:10.1101/2025.02.05.636714 , abstract =

    work page doi:10.1101/2025.02.05.636714 2025

  14. [14]

    Tahoe-100M: A Giga-Scale Single-Cell Perturbation Atlas for Context-Dependent Gene Function and Cellular Modeling.bioRxiv, 2025

    Zhang, Jesse and Ubas, Airol A. and Borja, Richard de and Svensson, Valentine and Thomas, Nicole and Thakar, Neha and Lai, Ian and Winters, Aidan and Khan, Umair and Jones, Matthew G. and Tran, Vuong and Pangallo, Joseph and Papalexi, Efthymia and Sapre, Ajay and Nguyen, Hoai and Sanderson, Oliver and Nigos, Maria and Kaplan, Olivia and Schroeder, Sarah a...

    work page doi:10.1101/2025.02.20.639398 2025

  15. [15]

    Cell Research , author =

    Grow. Cell Research , author =. 2025 , pmid =. doi:10.1038/s41422-025-01101-y , language =

    work page doi:10.1038/s41422-025-01101-y 2025

  16. [16]

    Nature , author =

    Where do proteins go in cells?. Nature , author =. 2025 , note =. doi:10.1038/d41586-025-01045-8 , abstract =

    work page doi:10.1038/d41586-025-01045-8 2025

  17. [17]

    Nature Cell Biology , author =

    Decoding heterogeneous single-cell perturbation responses , volume =. Nature Cell Biology , author =. 2025 , pmid =. doi:10.1038/s41556-025-01626-9 , abstract =

    work page doi:10.1038/s41556-025-01626-9 2025

  18. [18]

    Nature Biotechnology , author =

    Virtual cells for predictive immunotherapy , volume =. Nature Biotechnology , author =. 2025 , pmid =. doi:10.1038/s41587-025-02583-2 , language =

    work page doi:10.1038/s41587-025-02583-2 2025

  19. [19]

    doi:10.48550/arXiv.2501.06039 , abstract =

    Wenckstern, Johann and Jain, Eeshaan and Vasilev, Kiril and Pariset, Matteo and Wicki, Andreas and Gut, Gabriele and Bunne, Charlotte , month = jan, year =. doi:10.48550/arXiv.2501.06039 , abstract =

    work page doi:10.48550/arxiv.2501.06039

  20. [20]

    Replogle, Alexander N

    Transcriptome-wide analysis of differential expression in perturbation atlases , volume =. Nature Genetics , author =. 2025 , pmid =. doi:10.1038/s41588-025-02169-3 , abstract =

    work page doi:10.1038/s41588-025-02169-3 2025

  21. [21]

    doi:10.1101/2025.07.03.663009 , abstract =

    Zhang, Mingxuan and Swamy, Vinay and Cassius, Rowan and Dupire, Léo and Kanatsoulis, Charilaos and Paull, Evan and AlQuraishi, Mohammed and Karaletsos, Theofanis and Califano, Andrea , month = jul, year =. doi:10.1101/2025.07.03.663009 , abstract =

    work page doi:10.1101/2025.07.03.663009 2025

  22. [22]

    Cell , author =

    Perturb-. Cell , author =. 2025 , pmid =. doi:10.1016/j.cell.2025.05.022 , abstract =

    work page doi:10.1016/j.cell.2025.05.022 2025

  23. [23]

    2024 , pmid =

    Nature Biotechnology , author =. 2024 , pmid =. doi:10.1038/s41587-023-01772-1 , abstract =

    work page doi:10.1038/s41587-023-01772-1 2024

  24. [24]

    Cell , author =

    Human interpretable grammar encodes multicellular systems biology models to democratize virtual cell laboratories , volume =. Cell , author =. 2025 , pmid =. doi:10.1016/j.cell.2025.06.048 , abstract =

    work page doi:10.1016/j.cell.2025.06.048 2025

  25. [25]

    Adduri, Abhinav K. and Gautam, Dhruv and Bevilacqua, Beatrice and Imran, Alishba and Shah, Rohan and Naghipourfar, Mohsen and Teyssier, Noam and Ilango, Rajesh and Nagaraj, Sanjay and Dong, Mingze and Ricci-Tam, Chiara and Carpenter, Christopher and Subramanyam, Vishvak and Winters, Aidan and Tirukkovular, Sravya and Sullivan, Jeremy and Plosky, Brian S. ...

    work page doi:10.1101/2025.06.26.661135 2025

  26. [26]

    Nature Biomedical Engineering , author =

    A deep generative model for deciphering cellular dynamics and in silico drug discovery in complex diseases , issn =. Nature Biomedical Engineering , author =. 2025 , pmid =. doi:10.1038/s41551-025-01423-7 , abstract =

    work page doi:10.1038/s41551-025-01423-7 2025

  27. [27]

    URLhttps://doi.org/10.1038/s41592-025-02772-6

    Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines , volume =. Nature Methods , author =. 2025 , note =. doi:10.1038/s41592-025-02772-6 , abstract =

    work page doi:10.1038/s41592-025-02772-6 2025

  28. [28]

    doi:10.48550/arXiv.2508.02276 , abstract =

    Tang, Xiangru and Yu, Zhuoyun and Chen, Jiapeng and Cui, Yan and Shao, Daniel and Wang, Weixu and Wu, Fang and Zhuang, Yuchen and Shi, Wenqi and Huang, Zhi and Cohan, Arman and Lin, Xihong and Theis, Fabian and Krishnaswamy, Smita and Gerstein, Mark , month = aug, year =. doi:10.48550/arXiv.2508.02276 , abstract =

    work page doi:10.48550/arxiv.2508.02276

  29. [29]

    Michael and Arevalo, John and Bennion, Melissa and Boisseau, Nicolas and Borowa, Adriana and Boyd, Justin D

    Chandrasekaran, Srinivas Niranj and Ackerman, Jeanelle and Alix, Eric and Ando, D. Michael and Arevalo, John and Bennion, Melissa and Boisseau, Nicolas and Borowa, Adriana and Boyd, Justin D. and Brino, Laurent and Byrne, Patrick J. and Ceulemans, Hugo and Ch’ng, Carolyn and Cimini, Beth A. and Clevert, Djork-Arne and Deflaux, Nicole and Doench, John G. a...

    work page doi:10.1101/2023.03.23.534023 2023

  30. [30]

    Nature Methods , author =

    Cell. Nature Methods , author =. 2025 , pmid =. doi:10.1038/s41592-024-02528-8 , abstract =

    work page doi:10.1038/s41592-024-02528-8 2025

  31. [31]

    Li, Lanxiang and You, Yue and Liao, Wenyu and Fan, Xueying and Lu, Shihong and Cao, Ye and Li, Bo and Ren, Wenle and Fu, Yunlin and Kong, Jiaming and Zheng, Shuangjia and Chen, Jizheng and Liu, Xiaodong and Tian, Luyi , month = dec, year =. A. doi:10.1101/2024.12.23.630036 , abstract =

    work page doi:10.1101/2024.12.23.630036 2024

  32. [32]

    and Torkar, Michaela and Li, Donghui and Karaletsos, Theofanis , month = aug, year =

    Istrate, Ana-Maria and Milletari, Fausto and Castrotorres, Fabrizio and Tomczak, Jakub M. and Torkar, Michaela and Li, Donghui and Karaletsos, Theofanis , month = aug, year =. rbio1 - training scientific reasoning. doi:10.1101/2025.08.18.670981 , abstract =

    work page doi:10.1101/2025.08.18.670981 2025

  33. [33]

    Nature Methods , author =

    Morphological map of under- and overexpression of genes in human cells , volume =. Nature Methods , author =. 2025 , pmid =. doi:10.1038/s41592-025-02753-9 , abstract =

    work page doi:10.1038/s41592-025-02753-9 2025

  34. [34]

    Science Advances , author =

    Multiscale mathematical model-informed reinforcement learning optimizes combination treatment scheduling in glioblastoma evolution , volume =. Science Advances , author =. 2025 , pmid =. doi:10.1126/sciadv.adv3316 , abstract =

    work page doi:10.1126/sciadv.adv3316 2025

  35. [35]

    doi:10.1101/2025.09.15.676236 , abstract =

    Qin, Hua and Zhang, Yilin and Guo, Zhihan and Hu, Luni and Li, Qingsong and Cao, Lei and Xia, Tianyi and Deng, Ziqing and Zhang, Yong and Li, Yuxiang and Fang, Shuangsang , month = sep, year =. doi:10.1101/2025.09.15.676236 , abstract =

    work page doi:10.1101/2025.09.15.676236 2025

  36. [36]

    and Pruett-Miller, Shondra M

    Baysoy, Alev and Tian, Xiaolong and Zhang, Feifei and Renauer, Paul and Bai, Zhiliang and Shi, Hao and Li, Haikuo and Tao, Bo and Yang, Mingyu and Enninful, Archibald and Gao, Fu and Wang, Guangchuan and Zhang, Wanqiu and Tran, Thao and Patterson, Nathan Heath and Bao, Shuozhen and Dong, Chuanpeng and Xin, Shan and Zhong, Mei and Rankin, Sherri and Guy, C...

    work page doi:10.1101/2024.11.18.624106 2024

  37. [37]

    Nature Biotechnology , author =

    Disentanglement of single-cell data with biolord , volume =. Nature Biotechnology , author =. 2024 , pmid =. doi:10.1038/s41587-023-02079-x , abstract =

    work page doi:10.1038/s41587-023-02079-x 2024

  38. [38]

    Predicting cellular responses to complex perturbations in high-throughput screens.Molecular Systems Biology, 19(6):e11517, 2023

    Predicting cellular responses to complex perturbations in high-throughput screens , volume =. Molecular Systems Biology , author =. 2023 , pmid =. doi:10.15252/msb.202211517 , abstract =

    work page doi:10.15252/msb.202211517 2023

  39. [39]

    Alexander Wolf, and Fabian J

    Nature Methods , author =. 2019 , pmid =. doi:10.1038/s41592-019-0494-8 , abstract =

    work page doi:10.1038/s41592-019-0494-8 2019

  40. [40]

    and Polanski, Krzysztof and Eliasof, Moshe and Schonlieb, Carola-Bibiane and Teichmann, Sarah A

    Megas, Stathis and Chen, Daniel G. and Polanski, Krzysztof and Eliasof, Moshe and Schonlieb, Carola-Bibiane and Teichmann, Sarah A. , month = sep, year =. Celcomen: spatial causal disentanglement for single-cell and tissue perturbation modeling , shorttitle =. doi:10.48550/arXiv.2409.05804 , abstract =

    work page doi:10.48550/arxiv.2409.05804

  41. [41]

    Cell , author =

    Toward a foundation model of causal cell and tissue biology with a. Cell , author =. 2024 , pmid =. doi:10.1016/j.cell.2024.07.035 , language =

    work page doi:10.1016/j.cell.2024.07.035 2024

  42. [42]

    Nature Biomedical Engineering , author =

    Integrated in vivo combinatorial functional genomics and spatial transcriptomics of tumours to decode genotype-to-phenotype relationships , copyright =. Nature Biomedical Engineering , author =. 2025 , note =. doi:10.1038/s41551-025-01437-1 , abstract =

    work page doi:10.1038/s41551-025-01437-1 2025

  43. [43]

    2023 , note =

    Briefings in Bioinformatics , author =. 2023 , note =. doi:10.1093/bib/bbad338 , abstract =

    work page doi:10.1093/bib/bbad338 2023

  44. [44]

    doi:10.1101/2025.11.08.686890 , abstract =

    Lin, Xiang and Kong, Zhenglun and Ghosh, Soumya and Kellis, Manolis and Zitnik, Marinka , month = nov, year =. doi:10.1101/2025.11.08.686890 , abstract =

    work page doi:10.1101/2025.11.08.686890 2025

  45. [45]

    and Lord, Nathan D

    Ramjattun, Koushul and Wang, Alyson and Lee, Hannah and Giri, Shilpi and Chen, Yijia and MacDonald, William A. and Lord, Nathan D. and Poholek, Amanda C. and Lee, Youjin and Das, Jishnu , month = nov, year =. Characterizing spatial functional microniches with. doi:10.1101/2025.11.13.688264 , abstract =

    work page doi:10.1101/2025.11.13.688264 2025

  46. [46]

    Nature Communications , author =

    Prioritizing perturbation-responsive gene patterns using interpretable deep learning , volume =. Nature Communications , author =. 2025 , note =. doi:10.1038/s41467-025-61476-9 , abstract =

    work page doi:10.1038/s41467-025-61476-9 2025

  47. [47]

    Nature Biomedical Engineering , author =

    Identifying perturbations that boost. Nature Biomedical Engineering , author =. 2025 , note =. doi:10.1038/s41551-025-01357-0 , abstract =

    work page doi:10.1038/s41551-025-01357-0 2025

  48. [48]

    Teo, Alan Yue Yang and Gautier, Matthieu and Brock, Laurent and Tsai, Jennifer Y. J. and Coucy, Alexandra de and Laskaratos, Achilleas and Regazzi, Nicola and Barraud, Quentin and Sofroniew, Michael V. and Anderson, Mark A. and Courtine, Grégoire and Squair, Jordan W. and Skinnider, Michael A. , month = jun, year =. Identification of perturbation-responsi...

    work page doi:10.1101/2024.06.13.598641 2024

  49. [49]

    Cell , author =

    Spatial. Cell , author =. 2022 , pmid =. doi:10.1016/j.cell.2022.02.015 , abstract =

    work page doi:10.1016/j.cell.2022.02.015 2022

  50. [50]

    doi:10.1101/2024.09.27.615337 , abstract =

    Cao, Shenghao and Yang, Kaiyuan and Cheng, Jiabei and Li, Jiachen and Shen, Hong-Bin and Pan, Xiaoyong and Yuan, Ye , month = aug, year =. doi:10.1101/2024.09.27.615337 , abstract =

    work page doi:10.1101/2024.09.27.615337 2024

  51. [51]

    Squidiff: predicting cellular development and responses to perturbations using a diffusion model.Nature Methods, 2025

    Squidiff: predicting cellular development and responses to perturbations using a diffusion model , issn =. Nature Methods , author =. 2025 , pmid =. doi:10.1038/s41592-025-02877-y , abstract =

    work page doi:10.1038/s41592-025-02877-y 2025

  52. [52]

    Nature Biotechnology , author =

    Systema: a framework for evaluating genetic perturbation response prediction beyond systematic variation , issn =. Nature Biotechnology , author =. 2025 , pmid =. doi:10.1038/s41587-025-02777-8 , abstract =

    work page doi:10.1038/s41587-025-02777-8 2025

  53. [53]

    and Nassar, Marcel and Osiński, Błażej and Eksi, Ridvan and Yan, Zichao and Stark, Rory and Zhang, Kun and Graepel, Thore , month = oct, year =

    Wu, Yan and Wershof, Esther and Schmon, Sebastian M. and Nassar, Marcel and Osiński, Błażej and Eksi, Ridvan and Yan, Zichao and Stark, Rory and Zhang, Kun and Graepel, Thore , month = oct, year =. doi:10.48550/arXiv.2408.10609 , abstract =

    work page doi:10.48550/arxiv.2408.10609

  54. [54]

    Nature Computational Science , author =

    Scouter predicts transcriptional responses to genetic perturbations with large language model embeddings , copyright =. Nature Computational Science , author =. 2025 , note =. doi:10.1038/s43588-025-00912-8 , abstract =

    work page doi:10.1038/s43588-025-00912-8 2025

  55. [55]

    doi:10.1101/2025.07.08.663664 , abstract =

    Xu, Jing and Yang, Xiaolu and Li, Yunguang and Wang, Huan and Li, Yikai and Tang, Shijie and Guo, Shiwei and Zou, Jiawei and Shi, Xiaohan and Chen, Zexi and Zhong, Chaoliang and Li, Penghao and Jing, Wei and Zheng, Kailian and Wu, Xinqian and Gao, Dong and Chen, Luonan and Jin, Gang , month = jul, year =. doi:10.1101/2025.07.08.663664 , abstract =

    work page doi:10.1101/2025.07.08.663664 2025

  56. [56]

    Li, Krinos and Xiao, Xianglu and Deng, Shenglong and He, Lucas and Zhong, Zijun and Zou, Yuanjie and Zhan, Zhonghao and Hui, Zheng and Bao, Weiye and Yang, Guang , month = oct, year =. Large. doi:10.48550/arXiv.2510.07706 , abstract =

    work page doi:10.48550/arxiv.2510.07706

  57. [57]

    2025 , pmid =

    Nature Communications , author =. 2025 , pmid =. doi:10.1038/s41467-025-67481-2 , abstract =

    work page doi:10.1038/s41467-025-67481-2 2025

  58. [58]

    Nature Methods , author =

    Benchmarking algorithms for generalizable single-cell perturbation response prediction , issn =. Nature Methods , author =. 2025 , pmid =. doi:10.1038/s41592-025-02980-0 , abstract =

    work page doi:10.1038/s41592-025-02980-0 2025

  59. [59]

    and Yu, John and Merico, Daniele and Goodarzi, Hani and Alidoust, Nima , month = oct, year =

    Gandhi, Shreshth and Javadi, Farnoosh and Svensson, Valentine and Khan, Umair and Jones, Matthew G. and Yu, John and Merico, Daniele and Goodarzi, Hani and Alidoust, Nima , month = oct, year =. Tahoe-x1:. doi:10.1101/2025.10.23.683759 , abstract =

    work page doi:10.1101/2025.10.23.683759 2025

  60. [60]

    and Carpenter, Christopher and Prashar, Jaanak and Ricci-Tam, Chiara and Ilango, Rajesh and Teyssier, Noam and Konermann, Silvana and Hsu, Patrick D

    Youngblut, Nicholas D. and Carpenter, Christopher and Prashar, Jaanak and Ricci-Tam, Chiara and Ilango, Rajesh and Teyssier, Noam and Konermann, Silvana and Hsu, Patrick D. and Dobin, Alexander and Burke, David P. and Goodarzi, Hani and Roohani, Yusuf H. , month = mar, year =. doi:10.1101/2025.02.27.640494 , abstract =

    work page doi:10.1101/2025.02.27.640494 2025

  61. [61]

    and Buendia, Alejandro and Brunet, Anne and Zou, James , month = dec, year =

    Sun, Eric D. and Buendia, Alejandro and Brunet, Anne and Zou, James , month = dec, year =. doi:10.64898/2025.11.30.691355 , abstract =

    work page doi:10.64898/2025.11.30.691355 2025

  62. [62]

    Science , author =

    Massively multiplex chemical transcriptomics at single-cell resolution , volume =. Science , author =. 2020 , pmid =. doi:10.1126/science.aax6234 , abstract =

    work page doi:10.1126/science.aax6234 2020

  63. [63]

    Nature Communications , author =

    Multiplexed single-cell transcriptional response profiling to define cancer vulnerabilities and therapeutic mechanism of action , volume =. Nature Communications , author =. 2020 , pmid =. doi:10.1038/s41467-020-17440-w , abstract =

    work page doi:10.1038/s41467-020-17440-w 2020

  64. [64]

    and Zhou, Peijie and Wen, Han and Zhu, Yi and Guo, Tiannan , month = feb, year =

    Sun, Rui and Qian, Liujia and Li, Yongge and Cheng, Honghan and Xue, Zhangzhi and Zhang, Xuedong and Tan, Lingling and Zhan, Yuecheng and Hu, Wenbin and Xiao, Qi and Liu, Zhiwei and Zhang, Guangmei and Weinan, E. and Zhou, Peijie and Wen, Han and Zhu, Yi and Guo, Tiannan , month = feb, year =. A perturbation proteomics-based foundation model for virtual c...

    work page doi:10.1101/2025.02.07.637070 2025

  65. [65]

    and Hsieh, Tsung-Han S

    Huang, Ann C. and Hsieh, Tsung-Han S. and Zhu, Jiang and Michuda, Jackson and Teng, Ashton and Kim, Soohong and Rumsey, Elizabeth M. and Lam, Sharon K. and Anigbogu, Ikenna and Wright, Philip and Ameen, Mohamed and You, Kwontae and Graves, Christopher J. and Kim, Hyunsung John and Litterman, Adam J. and Sit, Rene V. and Blocker, Alex and Chu, Ci , month =...

    work page doi:10.1101/2025.06.11.659105 2025

  66. [66]

    2026 , file =

    arcinstitute/. 2026 , file =

    2026

  67. [67]

    Wei, Zhijian and Ma, Runze and Wang, Zichen and Li, Zhongmin and Song, Shuotong and Zheng, Shuangjia , month = oct, year =

  68. [68]

    doi:10.64898/2026.01.19.696785 , abstract =

    Zhang, Yuhui and Su, Yuchang and Wefers, Zoe and Su, Shiye and Li, He and Li, Tianhong and Wang, Chenyu and Burgess, James and Lozano, Alejandro and Zhou, Linqi and Ding, Daisy and Nirschl, Jeffrey and Lundberg, Emma and Yeung-Levy, Serena , month = jan, year =. doi:10.64898/2026.01.19.696785 , abstract =

    work page doi:10.64898/2026.01.19.696785 2026

  69. [69]

    Vascularized tumor organoids enable immunotherapy testing in tumor-remodeled stroma , copyright =

    Wimmer, Reiner and Krese, Rok and Henniger, Louisa and Niu, Steve and Lagarde, Nadege and Keller, Timon and Clausen, Inga and Köchel, Fabian and Koller, Claudia Ferrara and Waldhauer, Inja and Bacac, Marina and Gjorevski, Nikolche and Melnikov, Fjodor and Villenave, Remi , month = jan, year =. Vascularized tumor organoids enable immunotherapy testing in t...

    work page doi:10.64898/2026.01.23.701300 2026

  70. [70]

    Inferring virtual cell environments using multi-agent reinforcement learning , copyright =

    Kalafut, Noah Cohen and He, Chenfeng and Sheng, Jie and Chandrashekar, Pramod Bharadwaj and Choi, Jerome and Wang, Daifeng , month = nov, year =. Inferring virtual cell environments using multi-agent reinforcement learning , copyright =. doi:10.1101/2025.11.21.689815 , abstract =

    work page doi:10.1101/2025.11.21.689815 2025

  71. [71]

    Nature Reviews

    Human organoids as. Nature Reviews. Drug Discovery , author =. 2026 , pmid =. doi:10.1038/s41573-025-01317-y , abstract =

    work page doi:10.1038/s41573-025-01317-y 2026

  72. [72]

    Nature , author =

    Mapping convergent regulators of melanoma drug resistance by. Nature , author =. 2026 , note =. doi:10.1038/s41586-026-10367-0 , abstract =

    work page doi:10.1038/s41586-026-10367-0 2026

  73. [73]

    Cell , author =

    Global genetic interaction network of a human cell maps conserved principles and informs functional interpretation of gene co-essentiality profiles , volume =. Cell , author =. 2026 , note =. doi:10.1016/j.cell.2026.03.044 , abstract =

    work page doi:10.1016/j.cell.2026.03.044 2026

  74. [74]

    , month = feb, year =

    Dibaeinia, Payam and Babu, Sudarshan and Knudson, Mei and ElSheikh, Ali and Wen, Yibo and Liu, Han and Perera, Jason and Khan, Aly A. , month = feb, year =. Virtual. doi:10.64898/2026.02.04.703804 , abstract =

    work page doi:10.64898/2026.02.04.703804 2026

  75. [75]

    Cell , author =

    Deep learning of functional perturbations from condensate morphology , volume =. Cell , author =. 2026 , pmid =. doi:10.1016/j.cell.2026.05.010 , abstract =

    work page doi:10.1016/j.cell.2026.05.010 2026

  76. [76]

    PerturbDiff: Functional Diffusion for Single-Cell Perturbation Modeling.arXiv preprint arXiv:2602.19685, 2026

    Yuan, Xinyu and Liu, Xixian and Zhang, Ya Shi and Zhang, Zuobai and Guo, Hongyu and Tang, Jian , month = feb, year =. doi:10.48550/arXiv.2602.19685 , abstract =

    work page doi:10.48550/arxiv.2602.19685

  77. [77]

    Cell , volume =

    A Next Generation Connectivity Map: L1000 Platform and the First 1,000,000 Profiles , author =. Cell , volume =. 2017 , doi =

    2017

  78. [78]

    Nature Medicine , volume =

    The Drug Repurposing Hub: a Next-Generation Drug Library and Information Resource , author =. Nature Medicine , volume =. 2017 , doi =

    2017

  79. [79]

    Cell , volume =

    Mapping Information-Rich Genotype-Phenotype Landscapes with Genome-Scale Perturb-seq , author =. Cell , volume =. 2022 , doi =

    2022

  80. [80]

    Nature Methods , volume =

    scGPT: Toward Building a Foundation Model for Single-Cell Multi-Omics Using Generative AI , author =. Nature Methods , volume =. 2024 , doi =

    2024

Showing first 80 references.