Researchers at Karolinska Institutet and KTH have developed a computational method that can reveal how cells change and specialise in the body. The study, which has been published in the journal PNAS, can provide important knowledge about why this process sometimes goes wrong and leads to disease.
Cell differentiation is a fundamental process in the body. It enables stem cells to develop into different cell types, such as neurons in the brain or immune cells that guard against infection. When the process is disrupted it can lead to serious disease, but studying it is hard.
A fundamental challenge is that today's methods for analysing single cells destroy the cells when they are measured, which means that researchers only get a single snapshot in time. To address this problem, researchers at Karolinska Institutet and KTH have developed an algorithm called MultistageOT. It is based on mathematical principles known as optimal transport and can reconstruct the entire developmental process from a single snapshot of the cells' gene expression levels.
"When we sequence a cell, it is destroyed, and this means that we do not know what that cell would look like in the future. Our method makes it possible to model the entire developmental process, even if the cells are observed at a single time point," says Magnus Tronstad doctoral student at the Department of Medicine, Solna Karolinska Institutet.
Can predict how cells mature
The algorithm learns to fill out the missing intermediate stages of the development and can therefore predict how cells mature and what function they will have. In the study, the method was tested on data from blood cell development, a complex system where stem cells give rise to many different types of blood cells.
The results demonstrate that MultistageOT is capable not only of reconstructing developmental trajectories, but also of identifying cells that deviate from the expected process-an essential mechanism for avoiding spurious conclusions.
"This gives us a powerful tool to understand how cells make "decisions" about their future, which is central in understanding how diseases arise when differentiation goes wrong," says Joakim Dahlin docent at the same department at Karolinska Institutet.
The researchers emphasize that the method is general and can be used in different biological systems, even outside the animal kingdom.
The study is a collaboration between researchers at Karolinska Institutet and Professor Johan Karlsson at KTH. The study was financed by Vetenskapsrådet, Cancerfonden and Karolinska Institutet. The researchers declare no conflict of interest.
Publication
"MultistageOT: Multistage optimal transport infers trajectories from a snapshot of single-cell data" M. Tronstad, J. Karlsson, & J.S. Dahlin, Proc. Natl. Acad. Sci. U.S.A. 122 (50) e2516046122, online 11 December 2025, doi: 10.1073/pnas.2516046122
