How does an immature cell become a pigment cell, a nerve cell, or a blood cell? Modern single-cell methods can determine very precisely which genes are active in individual cells. From this, researchers can infer how cells change during development. What has so far only been possible to a limited extent, however, is predicting which regulatory genes drive this development – and what happens when individual regulatory genes are switched off.
A team led by Prof. Fabian Theis, Director of the Computational Health Center (CHC) at Helmholtz Munich and Professor at the TUM, as well as Tatjana Sauka-Spengler, Ph.D., Investigator at the Stowers Institute for Medical Research, has developed RegVelo, an AI-based model that closes this gap. RegVelo identifies from single-cell data which genes in a cell are currently becoming active or falling silent. At the same time, the model takes into account which genes regulate other genes. By combining data-driven learning with an explicit representation of gene-regulatory networks, RegVelo forms a hybrid model that links statistical inference with mechanistic insight. This means that RegVelo can not only trace developmental paths of cells; it can also simulate how these paths change when regulatory switches become active or stop acting.
RegVelo Considers Genes as Part of a Network
The approach builds on what is known as RNA velocity: single-cell data are used to estimate the direction in which a cell is likely to develop. This is based on a process that occurs in every active gene. First, immature RNA copies are produced, which are then further processed. The ratio of immature to processed RNA can be used to determine whether the activity of a gene is currently increasing or decreasing. Previous models considered this dynamic for each gene individually. RegVelo extends these approaches by integrating RNA velocity with gene-regulatory interactions, resulting in a hybrid, network-aware dynamical model.
“What made this work especially powerful was the combination of complementary strengths – high-resolution gene regulatory circuitry from our lab and dynamic trajectory and network modeling from Fabian Theis’s team” Sauka-Spengler said. “RegVelo emerged from integrating those two views into one framework.”
Weixu Wang, doctoral researcher at the CHC and first author of the study, describes the methodological challenge: “To take the network into account, we had to develop a mathematical model and carefully test whether it delivers robust predictions across different biological systems.” RegVelo was indeed able to reconstruct known developmental trajectories and regulators in several test systems – including the cell cycle, blood formation, and the development of pancreatic cells.
The researchers examined the development of neural crest cells in zebrafish in particular detail. During embryonic development, these cells give rise to pigment cells, nerve cells, and parts of the craniofacial tissue, among others. RegVelo identified already known regulators of cell development in zebrafish – and also pointed the researchers to previously less well-known candidates.
A Step Toward Virtual Cell Models
“RegVelo makes visible what consequences it has for a cell’s developmental path when a specific genetic regulator is switched off – and which genes and networks are involved,” says Weixu Wang. In this way, verifiable predictions can be derived from single-cell data, for example about which genetic regulators promote, slow down, or redirect a particular developmental path.
For Fabian Theis, the study points beyond this individual application: “RegVelo is a step toward virtual cell models that will help us better understand how cells behave in differentiation contexts and how they respond to genetic perturbation. Such hybrid models, combining data-driven AI with mechanistic biological structure, could be key to moving from description to prediction in biology.” This is initially basic research, but in the long term it could “help us better understand disease-relevant cell states and identify possible starting points for new therapies,” Theis concludes.
Original Publication
Wang et al., 2026: RegVelo: gene-regulatory-informed dynamics of single cells. Cell. DOI: 10.1016/j.cell.2026.04.022
More information
Read the press release from the Stowers Institute for Medical Research to learn more about the new paper, and watch the accompanying videos: https://www.stowers.org/news/regvelo-stowers
