Oxford RegVelo predicts cell fates

The University of Oxford and its collaborators said this week they had developed RegVelo, a computational framework designed to predict how single cells move toward specialised identities such as neurons, blood cells or pigment cells. Cell published the paper on May 11, and Oxford highlighted the work in a university news release on May 12. (ox.ac.uk) The researchers said the model combines RNA velocity — a way of estimating how gene expression is changing inside a cell — with gene regulatory networks that describe which genes control other genes. The team said that combination is intended to show not only where cells are headed, but which regulators may be steering them. (stowers.org) (github.com) ### What exactly did the researchers build? RegVelo is described by its authors as an end-to-end framework for inferring “regulatory cellular dynamics through coupled splicing dynamics.” The documentation says the software can estimate RNA velocity, infer latent time along a differentiation process, quantify velocity uncertainty and estimate the effects of perturbing gene regulons. The Cell paper describes it as a model that jointly learns splicing kinetics and gene regulatory relationships from single-cell gene expression data. The Oxford release said the framework was developed by researchers from Oxford, the Stowers Institute for Medical Research, Helmholtz Munich and the Technical University of Munich. (ox.ac.uk) The paper’s author list includes Weixu Wang, Zhiyuan Hu, Philipp Weiler, Sarah Mayes, Marius Lange, Jingye Wang, Zhengyuan Xue, Tatjana Sauka-Spengler and Fabian J. Theis. ### How is this different from standard RNA-velocity tools? The authors said existing RNA-velocity methods model how cells change over time but usually omit regulatory interactions between genes. They also said gene regulatory network methods tend to ignore the changing dynamics of cells. RegVelo was built to connect those two lines of analysis in one model, according to the paper and the Oxford and Stowers summaries. (regvelo.readthedocs.io) Fabian J. Theis and his co-authors wrote in the paper that the model is intended to be dynamic, interpretable and actionable. Tatjana Sauka-Spengler said in Oxford’s release that the goal is to understand “how cells make decisions and transition from one state to another,” and that RegVelo models how those decisions are encoded in gene regulatory networks over time. (ox.ac.uk) ### What evidence did they give that it works? The bioRxiv preprint and the Cell paper summary say the model was applied to datasets covering the cell cycle, human hematopoiesis and murine pancreatic endocrinogenesis. In those settings, the authors said RegVelo showed stronger predictive power for interactions and perturbation simulations than methods focused only on dynamics or only on regulatory-network inference. (biorxiv.org) Zebrafish neural crest cells were the main biological test case highlighted by Oxford and Stowers. The team said RegVelo identified tfec as an early driver of pigment cell formation and elf1 as a previously unknown regulator of pigment cell fate. The authors said those predictions were then tested with CRISPR-Cas9 knockout experiments and single-cell Perturb-seq. (biorxiv.org) ### Why are Oxford and the authors talking about regenerative medicine? Tatjana Sauka-Spengler said in the Stowers release that if researchers can identify the early instructions that direct cell fate, they may be able to reproduce some of those cell types in vitro for cell therapies in regenerative medicine. (biorxiv.org) Oxford’s release said the researchers believe the framework could accelerate work in developmental biology and regenerative medicine and improve understanding of disease-related cell states and therapeutic targets. Those are statements from the researchers and their institutions, not clinical results. The Stowers release also said the model could be relevant to developmental disorders and tumor growth. (ox.ac.uk) The published materials available here do not report a human clinical study or a medical product. They describe a research tool for analysing single-cell data and simulating how genetic changes could alter developmental pathways. ### Where can other researchers get it now? The code is already public. The RegVelo documentation says users can install the package from PyPI or from the GitHub repository, and the GitHub page shows the public repository under theislab/regvelo. A separate public repository, theislab/regvelo_reproducibility, contains notebooks to reproduce the paper’s results and says the datasets are available through a FigShare project. (stowers.org) The next step is likely to come through outside use of those materials. As of May 14, the documentation site, the code repository and the reproducibility repository were all online, and Oxford’s release points readers to the Cell paper for the full study record. (regvelo.readthedocs.io) (stowers.org)