AI Tool Predicts How Cells Choose Their Future

What are the first steps that chart the path for a cell to become a blood cell, neuron cell, or pigment cell? Scientists have developed increasingly powerful tools to track those changes, but one challenge has persisted: understanding not just where cells are headed, but which regulators steer them to their final fate. 

Now, new research from the Stowers Institute for Medical Research and Hemholtz Munich, published May 11, 2026, in Cell, has developed a new AI framework designed to help answer that question. RegVelo is a model that connects two areas of single-cell biology that have often remained separate: methods that estimate how cells change over time and methods that infer the gene regulatory networks controlling those changes.  

By bringing those pieces together, RegVelo allows researchers to time travel, predict how cells change, and identify which genes control those changes. They can do it all through computer simulations, eliminating the need to run every experiment in the lab.

“So, why is this important to know?” said Tatjana Sauka-Spengler, Ph.D., Stowers Institute Investigator and co-senior author of the study. “You can imagine, if you had a very early set of cells, being equipped with a particular set of instructions could allow you to reproduce, in vitro, some of these cell types in a very natural way. These cells could then be used in cell therapies in regenerative medicine.” 

“Sauka-Spengler and her collaborators have developed a meaningfully different way to process this kind of data,” said Stowers President and Chief Scientific Officer Alejandro Sánchez Alvarado, Ph.D. “It allows us to infer the most likely path of each component through space and time, and to use deep learning to predict those dynamics and test them experimentally.”

In the study, RegVelo modeled the neural crest, a group of early embryonic cells that can become many different parts of the body. In zebrafish neural crest development, RegVelo identified an early driver of pigment cell formation (tfec) and revealed a previously unknown regulator of pigment cell fate (elf1). Those predictions were then supported experimentally, showing that the model could do more than describe developmental change.

“There is always an initiating, driving element in something that will be defined at the end," said Sauka-Spengler. "But most times, if not always, that element is lost if you’re only analyzing that final cell state. Development is often described as a series of static snapshots of cell states. What we really want to understand, however, is how cells make decisions—how they transition from one state to another. RegVelo models how these fate decisions are encoded in gene regulatory networks over time and space, and what drives them.”

By helping connect early regulatory events to later cell fates, the work could also improve how scientists study developmental disorders and, over time, help guide efforts in regenerative medicine and cell therapy.

“RegVelo’s value extends well beyond neural crest cells,” said Sánchez Alvarado. “It’s applicable to any system in which cells change over time, from basic developmental biology to modeling tumor trajectories and the cellular outcomes that may inform treatment. It deserves attention from anyone working on cellular dynamics.”

Bridging a long-standing gap in single-cell biology

Single-cell biology research has made it possible to build increasingly detailed maps of development. RNA velocity methods can help researchers estimate how cells move through developmental landscapes, while gene regulatory network approaches can identify relationships among genes. But those methods have typically been used in parallel rather than together. RNA velocity methods often do not directly model transcriptional regulation, while regulatory network approaches generally do not capture cellular dynamics over time.  

“For a long time, cellular dynamics and gene regulation have largely been modeled separately,” said the study's co-senior author Prof. Fabian J. Theis, Ph.D., Director of the Computational Health Center (CHC) at Helmholtz Munich and Professor at the Technical University of Munich. “RegVelo brings those pieces together, allowing us to ask not only how cells are changing, but which regulatory interactions are helping drive those changes.”

The framework jointly models splicing kinetics and gene regulatory relationships, allowing researchers to map the hidden timeline of cell development, predict how cells shift from one state to another, and test what might happen when specific regulators are perturbed. In practical terms, that means scientists can ask a more mechanistic question than simply “Where is this cell going?” Now they can ask, “Which genes are helping push it there?”  

Joining forces

The work also reflects a deep collaboration between complementary teams. Sauka-Spengler's Lab, which transitioned from the University of Oxford to the Stowers Institute in 2022, brought a high-resolution gene regulatory scaffold for cranial neural crest development, while Theis’s group brought computational expertise in modeling and defining developmental trajectories of single cells and using RNA velocity analyses.

Together, those approaches were combined into a shared deep learning model that made developmental transitions highly predictive and testable. 

“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’s team,”  Sauka-Spengler said. "RegVelo emerged from integrating those two views into one framework.” 

Reference: Wang W, Hu Z, Weiler P, et al. RegVelo: Gene-regulatory-informed dynamics of single cells. Cell. 2026:S0092867426004575. doi: 10.1016/j.cell.2026.04.022

This article has been republished from the following materials. Note: material may have been edited for length and content. For further information, please contact the cited source. Our press release publishing policy can be accessed here.