Background
Much of what we know about biology comes from animal models, particularly mice. Applying these insights to humans, however, is not always straightforward. For example, some of the genes important for the formation of germ cells, which eventually give rise to eggs or sperm, differ between mice and humans. To identify gene correspondences across species, researchers often compare DNA sequences. This helps establish evolutionary relationships between genes, but it does not always answer a more functional question: which genes play similar regulatory roles within specific cell types or cell fate specification processes across species?
To explore this question, a team of researchers at Kyoto University’s Institute for the Advanced Study of Human Biology (WPI-ASHBi) looked at how genes are expressed in relation to one another. They reasoned that the overall organization of regulatory networks might remain similar across species, even when the genes involved differ. If so, this shared organization might be reflected in the geometry of gene expression as a whole. The team therefore developed Species-OT, a method that compares patterns of gene expression to identify genes that may play similar roles across species.
Methods and Findings
Species-OT compares transcriptome data from corresponding cell types or cell fate specification processes across species. Within each dataset, gene expression vectors can be regarded as points, placing genes with similar expression patterns closer together. These points form a geometric “shape,” characterizing the dataset. Then, using a mathematical theory called Gromov-Wasserstein optimal transport, the method compares how these points are arranged relative to one another to identify candidate gene correspondences across species. It also measures how much the overall gene expression patterns differ between datasets. In this study, the comparison focused on genes encoding transcription factors, which are expected to form the well-conserved core of gene regulatory networks.
To test the method, the researchers analyzed gene expression during the induction of primordial germ cell-like cells (PGCLCs) from mouse and human pluripotent stem cells. Among Species-OT’s top predictions were pairs of mouse and human genes that previous studies had shown to play similar roles. The analysis identified possible matches between genes that are closely related through evolution, as well as those that are more distantly related.
Species-OT also compares overall patterns of gene expression. In the analysis of PGCLC induction, these patterns were relatively similar in humans and cynomolgus macaques, while the mouse was more distant from both. The team further analyzed single-cell RNA-sequencing data from pluripotent stem cells of six mammalian species (humans, chimpanzees, gorillas, orangutans, cynomolgus macaques, and mice), which revealed similarities among primates, particularly between humans and chimpanzees, and among gorillas, orangutans, and cynomolgus macaques. In this analysis as well, mouse expression patterns differed substantially from those of the primates. These analyses reflect differences in cell type and culture conditions, which also affect gene expression, as well as differences between species. The resulting relationships are calculated differently from conventional phylogenetic trees.
Outlook
Species-OT is a data-driven method designed to infer gene correspondences across species, so that what is known about the role of a gene in one species can be carried over to its counterpart in another. It rests on the hypothesis that gene regulatory mechanisms shared across species are reflected in the “shape” of transcriptome data. Its predictions depend on how the input data are preprocessed and on how differences in gene expression are measured, and in this study the evaluation of gene correspondences focused mainly on a selected set of genes involved in germ cell specification. These findings provide a starting point for broader comparisons of gene regulation across species. Further work will involve testing the predicted gene correspondences experimentally and applying Species-OT to other cell types and cell fate specification processes, which is expected to advance understanding of which mechanisms are shared across species and which differ. Incorporating additional information about gene regulation could also extend the method and improve the accuracy of its predictions. Beyond comparisons between species, the same approach could be applied to differences between cell types, culture conditions, or individuals.
In the longer term, Species-OT could help translate findings from animal studies into a better understanding of human disease by identifying genes that may play similar roles across species. This could contribute to Moonshot Goal 2, which aims to realize a society capable of predicting and preventing diseases at an ultra-early stage by 2050. If the method can be applied to data from disease-relevant cells, it could help narrow down the human genes that correspond to changes found in animals and support studies testing which mechanisms are shared and which are specific to humans. The data examined in this study concern pluripotent stem cells and germ cell specification, however, and the value of the method for disease prediction and prevention remains to be tested.
“ I hope to use mathematical models and data science to understand the mathematical structures underlying complex natural phenomena, to control those phenomena, and to tackle the real-world problems associated with them,” says lead author Dr. Tokuta. “By comparing the mathematical structures underlying genomic, epigenomic, and gene expression data across species, this research could help us understand where and how humans differ from other organisms, as well as what we have in common.”
Glossary
Cell Reports Methods
Geometric inference of cross-species transcriptome correspondence
29-Oct-2026