{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,23]],"date-time":"2026-10-23T19:53:55Z","timestamp":1792785235268,"version":"4.5.0"},"reference-count":79,"publisher":"National Academy of Sciences","issue":"41","license":[{"start":{"date-parts":[[2026,10,8]],"date-time":"2026-10-08T00:00:00Z","timestamp":1791417600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"NCCR Automation","award":["51NF40 180545"],"award-info":[{"award-number":["51NF40 180545"]}]},{"name":"German Research Foundation","award":["EXC number 2064\/1"],"award-info":[{"award-number":["EXC number 2064\/1"]}],"id":[{"id":"https:\/\/ror.org\/018mejw64","id-type":"ROR","asserted-by":"crossref"}]},{"name":"Eric and Wendy Schmidt Center at the Broad Institute","award":["n\/a"],"award-info":[{"award-number":["n\/a"]}],"id":[{"id":"https:\/\/ror.org\/05a0ya142","id-type":"ROR","asserted-by":"crossref"}]},{"name":"NCCIH\/NIH","award":["1DP2AT012345"],"award-info":[{"award-number":["1DP2AT012345"]}]},{"name":"NIDDK\/NIH","award":["5RC2DK135492-02"],"award-info":[{"award-number":["5RC2DK135492-02"]}]},{"name":"ONR","award":["N00014-24-1-2687"],"award-info":[{"award-number":["N00014-24-1-2687"]}]},{"name":"United States Department of Energy","award":["DE-SC0023187"],"award-info":[{"award-number":["DE-SC0023187"]}],"id":[{"id":"https:\/\/ror.org\/01znn6x10","id-type":"ROR","asserted-by":"crossref"}]}],"content-domain":{"domain":["www.pnas.org"],"crossmark-restriction":true},"short-container-title":["Proc. Natl. Acad. Sci. U.S.A."],"published-print":{"date-parts":[[2026,10,13]]},"abstract":"<jats:p>Perturbation screens hold the potential to systematically map regulatory processes at single-cell resolution, yet modeling and predicting transcriptome-wide responses to perturbations remains a major computational challenge. Existing methods often underperform simple baselines, fail to disentangle measurement noise from biological signal, and provide limited insight into the causal structure governing cellular responses. Here, we present the latent causal diffusion (LCD), a generative model that frames single-cell gene expression as a stationary diffusion process observed under measurement noise. LCD outperforms established approaches in predicting the distributional shifts of unseen perturbation combinations in single-cell RNA-sequencing screens while simultaneously learning a mechanistic dynamical system of gene regulation. To interpret these learned dynamics, we develop an approach we call causal linearization via perturbation responses (CLIPR), which yields an approximation of the direct causal effects between all genes modeled by the diffusion. CLIPR provably identifies causal effects under a linear drift assumption and recovers causal structure in both simulated systems and a genome-wide perturbation screen, where it clusters genes into coherent functional modules and resolves causal relationships that standard differential expression analysis cannot. The LCD-CLIPR framework bridges generative modeling with causal inference to predict unseen perturbation effects and map the underlying regulatory mechanisms of the transcriptome.<\/jats:p>","DOI":"10.1073\/pnas.2602630123","type":"journal-article","created":{"date-parts":[[2026,10,8]],"date-time":"2026-10-08T15:42:01Z","timestamp":1791474121000},"update-policy":"https:\/\/doi.org\/10.1073\/pnas.cm10313","source":"Crossref","is-referenced-by-count":1,"title":["Latent causal diffusions for single-cell perturbation modeling"],"prefix":"10.1073","volume":"123","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-7465-5892","authenticated-orcid":false,"given":"Lars","family":"Lorch","sequence":"first","affiliation":[{"name":"Department of Computer Science, Eidgen\u00f6ssische Technische Hochschule (ETH) Z\u00fcrich","place":["Z\u00fcrich, Switzerland"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9039-6843","authenticated-orcid":false,"given":"Jiaqi","family":"Zhang","sequence":"additional","affiliation":[{"id":[{"id":"https:\/\/ror.org\/042nb2s44","id-type":"ROR","asserted-by":"publisher"}],"name":"Laboratory for Information and Decision Systems, Massachusetts Institute of Technology (MIT)","place":["Cambridge"]},{"id":[{"id":"https:\/\/ror.org\/05a0ya142","id-type":"ROR","asserted-by":"publisher"}],"name":"Eric and Wendy Schmidt Center, Broad Institute of MIT and Harvard","place":["Cambridge"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1431-103X","authenticated-orcid":false,"given":"Charlotte","family":"Bunne","sequence":"additional","affiliation":[{"id":[{"id":"https:\/\/ror.org\/02s376052","id-type":"ROR","asserted-by":"publisher"}],"name":"School of Computer and Communication Sciences, \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL)","place":["Lausanne, Switzerland"]},{"name":"School of Life Sciences, Swiss Institute for Experimental Cancer Research, EPFL","place":["Lausanne, Switzerland"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7260-9673","authenticated-orcid":false,"given":"Andreas","family":"Krause","sequence":"additional","affiliation":[{"name":"Department of Computer Science, Eidgen\u00f6ssische Technische Hochschule (ETH) Z\u00fcrich","place":["Z\u00fcrich, Switzerland"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8177-0925","authenticated-orcid":false,"given":"Bernhard","family":"Sch\u00f6lkopf","sequence":"additional","affiliation":[{"name":"Department of Computer Science, Eidgen\u00f6ssische Technische Hochschule (ETH) Z\u00fcrich","place":["Z\u00fcrich, Switzerland"]},{"id":[{"id":"https:\/\/ror.org\/04fq9j139","id-type":"ROR","asserted-by":"publisher"}],"name":"Max Planck Institute for Intelligent Systems","place":["T\u00fcbingen, Germany"]},{"name":"European Laboratory for Learning and Intelligent Systems (ELLIS) Institute","place":["T\u00fcbingen, Germany"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7008-0216","authenticated-orcid":false,"given":"Caroline","family":"Uhler","sequence":"additional","affiliation":[{"id":[{"id":"https:\/\/ror.org\/042nb2s44","id-type":"ROR","asserted-by":"publisher"}],"name":"Laboratory for Information and Decision Systems, Massachusetts Institute of Technology (MIT)","place":["Cambridge"]},{"id":[{"id":"https:\/\/ror.org\/05a0ya142","id-type":"ROR","asserted-by":"publisher"}],"name":"Eric and Wendy Schmidt Center, Broad Institute of MIT and Harvard","place":["Cambridge"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"341","published-online":{"date-parts":[[2026,10,8]]},"reference":[{"key":"e_1_3_4_1_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.cell.2022.05.013"},{"key":"e_1_3_4_2_2","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.2210623120"},{"key":"e_1_3_4_3_2","doi-asserted-by":"crossref","unstructured":"A. C. Huang et al. X-Atlas\/Orion: Genome-wide Perturb-seq datasets via a scalable Fix-Cryopreserve platform for training dose-dependent biological foundation models. bioRxiv [Preprint] (2025). https:\/\/doi.org\/10.1101\/2025.06.11.659105 (Accessed 11 June 2026).","DOI":"10.1101\/2025.06.11.659105"},{"key":"e_1_3_4_4_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.cell.2024.07.035"},{"key":"e_1_3_4_5_2","doi-asserted-by":"publisher","DOI":"10.1126\/science.aax4438"},{"key":"e_1_3_4_6_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41592-022-01705-x"},{"key":"e_1_3_4_7_2","unstructured":"B. Cleary A. Regev The necessity and power of random under-sampled experiments in biology. arXiv [Preprint] (2020). https:\/\/arxiv.org\/abs\/2012.12961 (Accessed 30 September 2025)."},{"key":"e_1_3_4_8_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41588-025-02124-2"},{"key":"e_1_3_4_9_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41556-023-01279-6"},{"key":"e_1_3_4_10_2","doi-asserted-by":"publisher","DOI":"10.1002\/cpt.3158"},{"key":"e_1_3_4_11_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41576-025-00920-4"},{"key":"e_1_3_4_12_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.cell.2024.11.015"},{"key":"e_1_3_4_13_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.csbj.2024.04.058"},{"key":"e_1_3_4_14_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41592-025-02772-6"},{"key":"e_1_3_4_15_2","unstructured":"T. Gaudelet et al. Season combinatorial intervention predictions with Salt & Peper. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2404.16907 (Accessed 30 September 2025)."},{"key":"e_1_3_4_16_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-023-06139-9"},{"key":"e_1_3_4_17_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41592-024-02201-0"},{"key":"e_1_3_4_18_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41592-024-02191-z"},{"key":"e_1_3_4_19_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41587-023-01905-6"},{"key":"e_1_3_4_20_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41592-023-01969-x"},{"key":"e_1_3_4_21_2","doi-asserted-by":"publisher","DOI":"10.15252\/msb.202211517"},{"key":"e_1_3_4_22_2","doi-asserted-by":"crossref","unstructured":"C. He J. Zhang M. Dahleh C. Uhler MORPH predicts the single-cell outcome of genetic perturbations across conditions and data modalities. bioRxiv [Preprint] (2025). https:\/\/doi.org\/10.1101\/2025.06.27.661992 (Accessed 10 April 2026).","DOI":"10.1101\/2025.06.27.661992"},{"key":"e_1_3_4_23_2","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.1714723115"},{"key":"e_1_3_4_24_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41576-021-00341-z"},{"key":"e_1_3_4_25_2","doi-asserted-by":"publisher","DOI":"10.1016\/S0006-3495(01)75949-8"},{"key":"e_1_3_4_26_2","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pbio.0040309"},{"key":"e_1_3_4_27_2","volume-title":"Stochastic Processes in Physics and Chemistry, North-Holland Personal Library","author":"Van Kampen N. G.","year":"2007","unstructured":"N. G. Van Kampen, Stochastic Processes in Physics and Chemistry, North-Holland Personal Library (Elsevier, ed. 3, 2007).","edition":"3"},{"key":"e_1_3_4_28_2","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.0803850105"},{"key":"e_1_3_4_29_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41592-018-0229-2"},{"key":"e_1_3_4_30_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.cels.2020.08.003"},{"key":"e_1_3_4_31_2","unstructured":"L. Lorch A. Krause B. Sch\u00f6lkopf \u201cCausal modeling with stationary diffusions\u201d in Proceedings of The 27th International Conference on Artificial Intelligence and Statistics Proceedings of Machine Learning Research S. Dasgupta S. Mandt Y. Li Eds. (Proceedings of Machine Learning Research (PMLR) 2024) vol. 238 pp. 1927\u20131935."},{"key":"e_1_3_4_32_2","doi-asserted-by":"publisher","DOI":"10.1126\/science.1070919"},{"key":"e_1_3_4_33_2","doi-asserted-by":"publisher","DOI":"10.1038\/ng869"},{"key":"e_1_3_4_34_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-642-14394-6"},{"key":"e_1_3_4_35_2","doi-asserted-by":"publisher","DOI":"10.1007\/BF02460618"},{"key":"e_1_3_4_36_2","doi-asserted-by":"publisher","DOI":"10.1038\/35002125"},{"key":"e_1_3_4_37_2","doi-asserted-by":"publisher","DOI":"10.1038\/nrg2102"},{"key":"e_1_3_4_38_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-021-23518-w"},{"key":"e_1_3_4_39_2","doi-asserted-by":"crossref","unstructured":"M. E. Vinyard et al. scDiffEq: Drift-diffusion modeling of single-cell dynamics with neural stochastic differential equations. bioRxiv (2023). https:\/\/doi.org\/10.1101\/2023.12.06.570508 (Accessed 2 December 2025).","DOI":"10.1101\/2023.12.06.570508"},{"key":"e_1_3_4_40_2","doi-asserted-by":"crossref","unstructured":"S. He et al. Squidiff: Predicting cellular development and responses to perturbations using a diffusion model. Nat. Methods 23 65\u201377 (2025).","DOI":"10.1038\/s41592-025-02877-y"},{"key":"e_1_3_4_41_2","doi-asserted-by":"publisher","DOI":"10.1093\/bib\/bbaf071"},{"key":"e_1_3_4_42_2","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/18.suppl_1.S136"},{"key":"e_1_3_4_43_2","doi-asserted-by":"publisher","DOI":"10.1038\/ng1165"},{"key":"e_1_3_4_44_2","doi-asserted-by":"publisher","DOI":"10.1038\/nature06847"},{"key":"e_1_3_4_45_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.cels.2016.04.001"},{"key":"e_1_3_4_46_2","doi-asserted-by":"publisher","DOI":"10.1016\/S0092-8674(00)00015-5"},{"key":"e_1_3_4_47_2","doi-asserted-by":"publisher","DOI":"10.1038\/nmeth.2930"},{"key":"e_1_3_4_48_2","first-page":"3","article-title":"A brief account of the theory of homogeneous Gaussian diffusions in finite dimensions","volume":"1","author":"Jacobsen M.","year":"1993","unstructured":"M. Jacobsen, A brief account of the theory of homogeneous Gaussian diffusions in finite dimensions. Front. Pure Appl. Probab. 1, 3 (1993).","journal-title":"Front. Pure Appl. Probab."},{"key":"e_1_3_4_49_2","first-page":"1","article-title":"Distinguishing cause from effect using observational data: Methods and benchmarks","volume":"17","author":"Mooij J. M.","year":"2016","unstructured":"J. M. Mooij, J. Peters, D. Janzing, J. Zscheischler, B. Sch\u00f6lkopf, Distinguishing cause from effect using observational data: Methods and benchmarks. J. Mach. Learn. Res. 17, 1\u2013102 (2016).","journal-title":"J. Mach. Learn. Res."},{"key":"e_1_3_4_50_2","doi-asserted-by":"crossref","unstructured":"L. Li et al. A systematic comparison of single-cell perturbation response prediction models. bioRxiv [Preprint] (2024). https:\/\/doi.org\/10.1101\/2024.12.23.630036 (Accessed 12 December 2025).","DOI":"10.1101\/2024.12.23.630036"},{"key":"e_1_3_4_51_2","first-page":"723","article-title":"A kernel two-sample test","volume":"13","author":"Gretton A.","year":"2012","unstructured":"A. Gretton, K. M. Borgwardt, M. J. Rasch, B. Sch\u00f6lkopf, A. Smola, A kernel two-sample test. J. Mach. Learn. Res. 13, 723\u2013773 (2012).","journal-title":"J. Mach. Learn. Res."},{"key":"e_1_3_4_52_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41587-025-02777-8"},{"key":"e_1_3_4_53_2","doi-asserted-by":"publisher","DOI":"10.1186\/1471-2105-14-128"},{"key":"e_1_3_4_54_2","doi-asserted-by":"crossref","unstructured":"A. Liberzon et al. Molecular signatures database (MSigDB) 3.0. Bioinformatics 27 1739\u20131740 (2011).","DOI":"10.1093\/bioinformatics\/btr260"},{"key":"e_1_3_4_55_2","doi-asserted-by":"publisher","DOI":"10.1111\/j.2517-6161.1995.tb02031.x"},{"key":"e_1_3_4_56_2","doi-asserted-by":"crossref","unstructured":"S. Brin R. Motwani J. D. Ullman S. Tsur \u201cDynamic itemset counting and implication rules for market basket data\u201d in Proceedings of the 1997 ACM SIGMOD International Conference on Management of Data J. M. Peckman S. Ram M. Franklin Eds. (Association for Computing Machinery 1997) pp. 255\u2013264.","DOI":"10.1145\/253260.253325"},{"key":"e_1_3_4_57_2","doi-asserted-by":"publisher","DOI":"10.1111\/rssb.12167"},{"key":"e_1_3_4_58_2","unstructured":"K. Yang A. Katcoff C. Uhler \u201cCharacterizing and learning equivalence classes of causal DAGs under interventions\u201d in International Conference on Machine Learning (PMLR) J. Dy A. Krause Eds. (2018) pp. 5541\u20135550."},{"key":"e_1_3_4_59_2","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2186"},{"key":"e_1_3_4_60_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41580-021-00335-z"},{"key":"e_1_3_4_61_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.cell.2024.04.035"},{"key":"e_1_3_4_62_2","doi-asserted-by":"publisher","DOI":"10.18632\/oncotarget.16723"},{"key":"e_1_3_4_63_2","doi-asserted-by":"publisher","DOI":"10.1126\/science.aax6234"},{"key":"e_1_3_4_64_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.cell.2018.05.061"},{"key":"e_1_3_4_65_2","unstructured":"H. E. Robbins \u201cAn empirical Bayes approach to statistics\u201d in Proceedings of the Third Berkeley Symposium on Mathematical Statistics and Probability (1954\/1955) Vol. I J. Neyman Ed. (University of California Press 1956) pp. 157\u2013163."},{"key":"e_1_3_4_66_2","doi-asserted-by":"publisher","DOI":"10.1111\/j.2517-6161.1984.tb01290.x"},{"key":"e_1_3_4_67_2","doi-asserted-by":"publisher","DOI":"10.1162\/NECO_a_00142"},{"key":"e_1_3_4_68_2","unstructured":"Y. Song S. Ermon \u201cGenerative modeling by estimating gradients of the data distribution\u201d in Advances in neural information processing systems H. Wallach et al. Eds. (Curran Associates Inc. 2019) vol. 32."},{"key":"e_1_3_4_69_2","doi-asserted-by":"publisher","DOI":"10.1017\/9781108186735"},{"key":"e_1_3_4_70_2","unstructured":"Y. Song . \u201cScore-based generative modeling through stochastic differential equations\u201d in International Conference on Learning Representations (2021)."},{"key":"e_1_3_4_71_2","unstructured":"V. Dumoulin J. Shlens M. Kudlur \u201cA learned representation for artistic style\u201d in International Conference on Learning Representations (2017)."},{"key":"e_1_3_4_72_2","unstructured":"A. Vaswani et al. \u201cAttention is all you need\u201d in Advances in neural information processing systems I. Guyon et al. Eds. (Curran Associates Inc. 2017) vol. 30."},{"key":"e_1_3_4_73_2","unstructured":"D. Kingma J. Ba \u201cAdam: A method for stochastic optimization\u201d in International Conference on Learning Representations (2015)."},{"key":"e_1_3_4_74_2","doi-asserted-by":"publisher","DOI":"10.2140\/involve.2014.7.751"},{"key":"e_1_3_4_75_2","doi-asserted-by":"publisher","DOI":"10.1038\/nbt.3192"},{"key":"e_1_3_4_76_2","first-page":"290","article-title":"On random graphs","volume":"6","author":"Erd\u0151s P.","year":"1959","unstructured":"P. Erd\u0151s, A. R\u00e9nyi, On random graphs. Publ. Math. 6, 290\u2013297 (1959).","journal-title":"Publ. Math."},{"key":"e_1_3_4_77_2","doi-asserted-by":"publisher","DOI":"10.1126\/science.286.5439.509"},{"key":"e_1_3_4_78_2","doi-asserted-by":"publisher","DOI":"10.1023\/A:1010920819831"},{"key":"e_1_3_4_79_2","unstructured":"L. Lorch et al. \u201cCode for: Latent causal diffusions for single-cell perturbation modeling.\u201d GitHub. https:\/\/github.com\/larslorch\/lcd-clipr. Deposited 24 September 2026."}],"container-title":["Proceedings of the National Academy of Sciences"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/pnas.org\/doi\/pdf\/10.1073\/pnas.2602630123","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/pnas.org\/doi\/pdf\/10.1073\/pnas.2602630123","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,8]],"date-time":"2026-10-08T15:42:16Z","timestamp":1791474136000},"score":1,"resource":{"primary":{"URL":"https:\/\/pnas.org\/doi\/10.1073\/pnas.2602630123"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,10,8]]},"references-count":79,"journal-issue":{"issue":"41","published-print":{"date-parts":[[2026,10,13]]}},"alternative-id":["10.1073\/pnas.2602630123"],"URL":"https:\/\/doi.org\/10.1073\/pnas.2602630123","relation":{},"ISSN":["0027-8424","1091-6490"],"issn-type":[{"value":"0027-8424","type":"print"},{"value":"1091-6490","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,10,8]]},"assertion":[{"value":"2026-01-23","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-09-01","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-10-08","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e2602630123"}}