Generative models of cell dynamics: from Neural ODEs to flow matching

Summarize this content to 100 words: Newman, S. A. Cell differentiation: what have we learned in 50 years? J. Theor. Biol. 485, 110031 (2020).
Google Scholar 
Bergen, V., Lange, M., Peidli, S., Wolf, F. A. & Theis, F. J. Generalizing rna velocity to transient cell states through dynamical modeling. Nat. Biotechnol. 38, 1408–1414 (2020).
Google Scholar 
Shi, J., Aihara, K., Li, T. & Chen, L. Energy landscape decomposition for cell differentiation with proliferation effect. Natl. Sci. Rev. 9, nwac116 (2022).
Google Scholar 
Waddington, C. H. The Strategy of the Genes, 1st edn. https://doi.org/10.4324/9781315765471. Reprinted as eBook in 2014. (Routledge, London, 1957).Qiu, X., Ding, S. & Shi, T. From understanding the development landscape of the canonical fate-switch pair to constructing a dynamic landscape for two-step neural differentiation. PLoS ONE 7, e49271 (2012).
Google Scholar 
Weinreb, C., Wolock, S., Tusi, B. K., Socolovsky, M. & Klein, A. M. Fundamental limits on dynamic inference from single-cell snapshots. Proc. Natl. Acad. Sci. 115, E2467–E2476 (2018).
Google Scholar 
Angerer, P. et al. Single cells make big data: new challenges and opportunities in transcriptomics. Curr. Opin. Syst. Biol. 4, 85–91 (2017).
Google Scholar 
Qiu, X. et al. Mapping transcriptomic vector fields of single cells. Cell 185, 690–711 (2022).
Google Scholar 
Teschendorff, A. E. & Feinberg, A. P. Statistical mechanics meets single-cell biology. Nat. Rev. Genet. 22, 459–476 (2021).
Google Scholar 
Deconinck, L., Cannoodt, R., Saelens, W., Deplancke, B. & Saeys, Y. Recent advances in trajectory inference from single-cell omics data. Curr. Opin. Syst. Biol. 27, 100344 (2021).
Google Scholar 
Trapnell, C. et al. The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. Nat. Biotechnol. 32, 381–386 (2014).
Google Scholar 
Qiu, X. et al. Inferring causal gene regulatory networks from coupled single-cell expression dynamics using scribe. Cell Syst. 10, 265–274 (2020).
Google Scholar 
Aliee, H. et al. Sparsity in continuous-depth neural networks. Adv. Neural Inf. Process. Syst. 35, 901–914 (2022). Systematically explores the impact of weight and feature sparsity on the generalization and dynamical law identification capabilities of Neural Ordinary Differential Equations (NODEs), introducing a novel regularization method, PathReg, to enhance input-output relation sparsity, and evaluate it across real-world datasets to assess out-of-distribution generalization in dynamical systems.
Google Scholar 
Gayoso, A. et al. Deep generative modeling of transcriptional dynamics for RNA velocity analysis in single cells. Nat. Methods 21, 50–59 (2024).Chen, Z., King, W. C., Hwang, A., Gerstein, M. & Zhang, J. Deepvelo: single-cell transcriptomic deep velocity field learning with neural ordinary differential equations. Sci. Adv. 8, eabq3745 (2022).
Google Scholar 
Lange, M. et al. Cellrank for directed single-cell fate mapping. Nat. methods 19, 159–170 (2022).
Google Scholar 
Weiler, P., Lange, M., Klein, M., Pe’er, D. & Theis, F. Unified fate mapping in multiview single-cell data. Nat. Methods 21, 1196–1205 (2024).Klein, D. et al. Mapping cells through time and space with moscot. Nature 638, 1065–1075 (2025).Lipman, Y., Chen, R. T. Q., Ben-Hamu, H., Nickel, M. & Le, M. Flow matching for generative modeling. In The Eleventh International Conference on Learning Representations. https://openreview.net/forum?id=PqvMRDCJT9t (2023).Tong, A. et al. Improving and generalizing flow-based generative models with minibatch optimal transport. Trans. Mach. Learn. Res. https://openreview.net/forum?id=CD9Snc73AW. Expert Certification (2024).Trapnell, C., Cacchiarelli, D. & Qiu, X. Monocle: Cell counting, differential expression, and trajectory analysis for single-cell RNA-seq experiments. Bioconductor https://www.bioconductor.org/packages/release/bioc/html/monocle.html (2017).Saunders, L. M. et al. Embryo-scale reverse genetics at single-cell resolution. Nature 623, 782–791 (2023).Street, K. et al. Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics. BMC Genom. 19, 1–16 (2018).
Google Scholar 
Ji, Z. & Ji, H. Tscan: Pseudo-time reconstruction and evaluation in single-cell RNA-seq analysis. Nucleic Acids Res. 44, e117–e117 (2016).
Google Scholar 
Setty, M. et al. Characterization of cell fate probabilities in single-cell data with Palantir. Nat. Biotechnol. 37, 451–460 (2019).
Google Scholar 
Haghverdi, L., Büttner, M., Wolf, F. A., Buettner, F. & Theis, F. J. Diffusion pseudotime robustly reconstructs lineage branching. Nat. methods 13, 845–848 (2016).
Google Scholar 
Wei, J., Zhou, T., Zhang, X. & Tian, T. Dtflow: inference and visualization of single-cell pseudotime trajectory using diffusion propagation. Genom. Proteom. Bioinform. 19, 306–318 (2021).
Google Scholar 
Mao, Q., Wang, L., Goodison, S. & Sun, Y. Dimensionality reduction via graph structure learning. In Proc. 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 765–774 (Association for Computing Machinery, New York, NY, 2015).Saelens, W., Cannoodt, R., Todorov, H. & Saeys, Y. A comparison of single-cell trajectory inference methods. Nat. Biotechnol. 37, 547–554 (2019).
Google Scholar 
Farrell, S., Mani, M. & Goyal, S. Inferring single-cell transcriptomic dynamics with structured latent gene expression dynamics. Cell Rep. Methods 3, 100581 (2023). Available at SSRN 4330809 LatentVelo leverages deep learning to compute low-dimensional gene dynamics, embedding cells in a dynamics-based latent space via a variational autoencoder and modeling differentiation with neural ordinary differential equations, enabling accurate prediction of cell trajectories and superior batch correction by focusing on gene expression dynamics.Ocone, A., Haghverdi, L., Mueller, N. S. & Theis, F. J. Reconstructing gene regulatory dynamics from high-dimensional single-cell snapshot data. Bioinformatics 31, i89–i96 (2015).
Google Scholar 
La Manno, G. et al. RNA velocity of single cells. Nature 560, 494–498 (2018).
Google Scholar 
Gu, Y., Blaauw, D. & Welch, J. D. Bayesian inference of RNA velocity from multi-lineage single-cell data. bioRxiv https://doi.org/10.1101/2022.07.08.499381 (2022).Wang, W. et al. Regvelo: gene-regulatory-informed dynamics of single cells. bioRxiv https://doi.org/10.1101/2024.12.11.627935 (2024).Granger, C. W. Some recent development in a concept of causality. J. Econ. 39, 199–211 (1988).
Google Scholar 
Schölkopf, B. et al. Toward causal representation learning. Proc. IEEE 109, 612–634 (2021).
Google Scholar 
Aliee, H., Theis, F. J. & Kilbertus, N. Beyond predictions in neural odes: identification and interventions. arXiv preprint https://doi.org/10.48550/arXiv.2106.12430 (2021).Tejada-Lapuerta, A. et al. Causal machine learning for single-cell genomics. Nat. Genet. 57, 797–808 (2025).Eraslan, G., Simon, L. M., Mircea, M., Mueller, N. S. & Theis, F. J. Single-cell RNA-seq denoising using a deep count autoencoder. Nat. Commun. 10, 390 (2019).
Google Scholar 
Lopez, R., Regier, J., Cole, M. B., Jordan, M. I. & Yosef, N. Deep generative modeling for single-cell transcriptomics. Nat. methods 15, 1053–1058 (2018).
Google Scholar 
Chen, R. T. Q., Rubanova, Y., Bettencourt, J. & Duvenaud, D. K. Neural Ordinary Differential Equations. In Advances in Neural Information Processing Systems Vol. 31 (Curran Associates, Inc., 2018).Mooij, J. M., Janzing, D. & Schölkopf, B. From ordinary differential equations to structural causal models: the deterministic case. In Proceedings of the Twenty-Ninth Conference on Uncertainty in Artificial Intelligence. UAE’13, pp-440–448. https://doi.org/10.48550/arXiv.1304.7920 (AUAI Press, Bellevue, WA, 2013).Chen, R. T. Q. torchdiffeq. https://github.com/rtqichen/torchdiffeq (2018).Tong, A., Huang, J., Wolf, G., Van Dijk, D. & Krishnaswamy, S. Trajectorynet: A dynamic optimal transport network for modeling cellular dynamics. In International conference on machine learning, 9526–9536 (PMLR, 2020). TrajectoryNet leverages the integration of continuous normalizing flows with dynamic optimal transport to model continuous, non-linear paths in dynamic systems, overcoming the limitations of static optimal transport models by constraining the flow of entities between distributions, showcasing enhanced modeling of cellular dynamics in scRNA-seq data.Hossain, I., Fanfani, V., Fischer, J., Quackenbush, J. & Burkholz, R. Biologically informed neuralodes for genome-wide regulatory dynamics. Genome Biol. 25, 127 (2024). PHOENIX advances ODE-based modeling of gene expression dynamics by integrating neural ODEs with Hill-Langmuir kinetics, offering a scalable framework that enhances biological interpretability through sparse representations and prior knowledge incorporation, demonstrated across in silico experiments, yeast cell oscillations, and genome-scale breast cancer studies.
Google Scholar 
Bertin, P. et al. A scalable gene network model of regulatory dynamics in single cells. arXiv preprint arXiv https://doi.org/10.48550/arXiv.2503.20027 (2025).Li, Q. sctour: a deep learning architecture for robust inference and accurate prediction of cellular dynamics. Genome Biol. 24, 149 (2023).
Google Scholar 
Schiebinger, G. et al. Optimal-transport analysis of single-cell gene expression identifies developmental trajectories in reprogramming. Cell 176, 928–943 (2019).
Google Scholar 
Lange, M. et al. Mapping lineage-traced cells across time points with Moslin. Genome Biol. 25, 277 (2024).
Google Scholar 
Eyring, L. et al. Unbalancedness in neural monge maps improves unpaired domain translation. The Twelfth International Conference on Learning Representations. https://openreview.net/forum?id=2UnCj3jeao (2024).Bunne, C., Papaxanthos, L., Krause, A. & Cuturi, M. Proximal optimal transport modeling of population dynamics. In International Conference on Artificial Intelligence and Statistics, 6511–6528 (PMLR, 2022).Benamou, J.-D. & Brenier, Y. A computational fluid mechanics solution to the monge-kantorovich mass transfer problem. Numer. Math. 84, 375–393 (2000).
Google Scholar 
Neklyudov, K., Brekelmans, R., Severo, D. & Makhzani, A. Action matching: learning stochastic dynamics from samples. In International conference on machine learning, 25858–25889 (PMLR, 2023).Huguet, G. et al. Manifold interpolating optimal-transport flows for trajectory inference. In Advances in Neural Information Processing Systems. Vol. 35, (eds Koyejo, S. et al.) 29705–29718 (Curran Associates, Inc., 2022). MIOFlow introduces a novel integration of dynamic models, manifold learning, and optimal transport through Neural ODEs and a geodesic autoencoder, offering superior interpolation between static population snapshots by embedding data in a geometry-aware latent space, outperforming existing generative models in capturing continuous population dynamics.Koshizuka, T. & Sato, I. Neural lagrangian schrödinger bridge: Diffusion modeling for population dynamics. In The Eleventh International Conference on Learning Representations (2023). Addresses the challenge of modeling biological population dynamics with coarse temporal data by introducing a method that combines the Lagrangian Schrödinger bridge problem with neural stochastic differential equations, enabling the approximation of both deterministic and stochastic sample trajectories in line with the principle of least action, and demonstrating efficient high-dimensional dynamics estimation.Bunne, C., Hsieh, Y.-P., Cuturi, M. & Krause, A. The schrödinger bridge between Gaussian measures has a closed form. In International Conference on Artificial Intelligence and Statistics, 5802–5833 (PMLR, 2023).Sha, Y., Qiu, Y., Zhou, P. & Nie, Q. Reconstructing growth and dynamic trajectories from single-cell transcriptomics data. Nat. Mach. Intell. 6, 25–39 (2024).
Google Scholar 
Palma, A. et al. Generating multi-modal and multi-attribute single-cell counts with CFGen. In The Thirteenth International Conference on Learning Representations. https://openreview.net/forum?id=3MnMGLctKb (2025).Tong, A. et al. Simulation-free schrödinger bridges via score and flow matching. In Proceedings of The 27th International Conference on Artificial Intelligence and Statistics. https://proceedings.mlr.press/v238/tong24a.html (2023).Klein, D., Uscidda, T., Theis, F. & Cuturi, M. Generative entropic neural optimal transport to map within and across spaces. Advances. in Neural Information Processing Systems. Vol. 37, 103897–103944 https://proceedings.neurips.cc/paper_files/paper/2024/file/bc46e29f91e676747c584ca181cb0ea1-Paper-Conference.pdf (2024).Neklyudov, K. et al. A computational framework for solving Wasserstein Lagrangian flows. In Proceedings of the 41st International Conference on Machine Learning. https://doi.org/10.48550/arXiv.2310.10649 (2023).Kapusniak, K. et al. Metric flow matching for smooth interpolations on the data manifold. In The Thirty-eighth Annual Conference on Neural Information Processing Systems. https://doi.org/10.48550/arXiv.2405.14780 (2024).Maizels, R. J., Snell, D. M. & Briscoe, J. Reconstructing developmental trajectories using latent dynamical systems and time-resolved transcriptomics. Cell Syst. 15, 411–424 (2024).
Google Scholar 
Zhang, Z., Wang, Z., Sun, Y., Li, T. & Zhou, P. Modeling cell dynamics and interactions with unbalanced mean field Schrödinger bridge. In The Thirty-ninth Annual Conference on Neural Information Processing Systems. https://doi.org/10.48550/arXiv.2505.11197 (2025).Krumsiek, J., Marr, C., Schroeder, T. & Theis, F. J. Hierarchical differentiation of myeloid progenitors is encoded in the transcription factor network. PloS one 6, e22649 (2011).
Google Scholar 
Trimbour, R., Deutschmann, I. M. & Cantini, L. Molecular mechanisms reconstruction from single-cell multi-omics data with hummus. Bioinformatics 40, btae143 (2024).Bergen, V., Soldatov, R. A., Kharchenko, P. V. & Theis, F. J. RNA velocity-current challenges and future perspectives. Mol. Syst. Biol. 17, e10282 (2021).
Google Scholar 
Brooks, T. et al. Video generation models as world simulators. 2024. https://openai.com/research/video-generation-models-as-world-simulators3 (2024).Zhu, Z. et al. Is Sora a world simulator? A comprehensive survey on general world models and beyond. arXiv preprint https://doi.org/10.48550/arXiv.2405.03520 (2024).Szałata, A. et al. Transformers in single-cell omics: a review and new perspectives. Nat. Methods 21, 1430–1443 (2024).
Google Scholar 
Liu, R., Pisco, A. O., Braun, E., Linnarsson, S. & Zou, J. Dynamical systems model of RNA velocity improves inference of single-cell trajectory, pseudo-time and gene regulation. J. Mol. Biol. 434, 167606 (2022).
Google Scholar 
Yeo, G. H. T., Saksena, S. D. & Gifford, D. K. Generative modeling of single-cell time series with prescient enables prediction of cell trajectories with interventions. Nat. Commun. 12, 3222 (2021).
Google Scholar 
Gayoso, A. et al. Deep generative modeling of transcriptional dynamics for RNA velocity analysis in single cells. Nat. Methods 21, 50–59 (2022).Pfister, N., Bauer, S. & Peters, J. Learning stable and predictive structures in kinetic systems. Proc. Natl. Acad. Sci. 116, 25405–25411 (2019).
Google Scholar 
Yuan, B. et al. Cellbox: interpretable machine learning for perturbation biology with application to the design of cancer combination therapy. Cell Syst. 12, 128–140 (2021).
Google Scholar 
Qian, Z., Kacprzyk, K. & van der Schaar, M. D-code: Discovering closed-form odes from observed trajectories. In International Conference on Learning Representations (ICLR, 2022).Kidger, P., Morrill, J., Foster, J. & Lyons, T. Neural controlled differential equations for irregular time series. In Advances in Neural Information Processing Systems. Vol. 33 (eds Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M. & Lin, H.) 6696–6707 (Neural Information Processing Systems Foundation, Inc. NeurIPS, 2020).Qian, Z., Zame, W., Fleuren, L., Elbers, P. & van der Schaar, M. Integrating expert odes into neural odes: pharmacology and disease progression. Adv. Neural Inf. Process. Syst. 34, 11364–11383 (2021).
Google Scholar 
Salvador, M. et al. Real-time whole-heart electromechanical simulations using latent neural ordinary differential equations. npj Digit. Med. 7, 90 (2024).Bellot, A., Branson, K. & van der Schaar, M. Neural graphical modelling in continuous-time: consistency guarantees and algorithms. In International Conference on Learning Representations (ICLR, 2022).Lu, J., Deng, K., Zhang, X., Liu, G. & Guan, Y. Neural-ode for pharmacokinetics modeling and its advantage to alternative machine learning models in predicting new dosing regimens. iScience 24, 102804 (2021).
Google Scholar 
Seedat, N., Imrie, F., Bellot, A., Qian, Z. & van der Schaar, M. Continuous-time modeling of counterfactual outcomes using neural controlled differential equations. In Proc. 39th International Conference on Machine Learning. Vol. 162, 19497–19521 (PMLR, 2022).



Source link