Universal cell embedding provides a foundation model for cell biology
7 affiliations
- 1Department of Computer Science, Stanford University, Stanford, CA, USA
- 2Department of Biomedical Data Science, Stanford University, Stanford, CA, USA
- 3Biohub, San Francisco, CA, USA. [email protected]
- 4Chan Zuckerberg Initiative, Redwood City, CA, USA. [email protected]
- 5Department of Applied Physics, Stanford University, Stanford, CA, USA. [email protected]
- 6Department of Bioengineering, Stanford University, Stanford, CA, USA. [email protected]
- 7Department of Computer Science, Stanford University, Stanford, CA, USA. [email protected]
Abstract
Developing a universal representation space for cells that encompasses the tremendous molecular diversity of cell types across species would be transformative for cell biology. Recent work using single-cell transcriptomic approaches to create molecular definitions of cell types in the form of cell atlases has provided the necessary data for such an endeavour1–3. Here we present the universal cell embedding (UCE) foundation model. UCE was trained on a large corpus of cell data using self-supervision, creating a unified biological latent space that can represent cells across diverse tissues and species. This latent space captures important biological variation despite the presence of experimental noise. UCE’s universality means that new cells can be embedded with no data labelling, model training or fine-tuning. We used UCE to create the Integrated Mega-scale Atlas, embedding 36 million cells, with more than 1,000 uniquely named cell types, from hundreds of experiments, dozens of tissues and eight species. We gain insights into the organization of cell types and tissues within the space. UCE’s embedding space exhibits emergent behaviour, identifying biology that it was never trained for, such as identifying developmental lineages and embedding data from species that were not included in the training set. Overall, by enabling a universal representation for every cell state and type, UCE is a valuable tool for analysis, annotation and hypothesis generation over single-cell data. The universal cell embedding foundation model learns to capture the organization and variation of cells by training on 36 million cells from hundreds of experiments, dozens of tissues and eight species.
Related
- Draft hybrid genome assembly of the photosynthetic sacoglossan Elysia crispata provides insights into sacoglossan biology.
- Spatial Metabolomics-Based Integrated Analysis Provides Evidence for the Mechanism of Salvianolic Acid A-Induced Mitochondrial Apoptosis in Non-Small Cell Lung Cancer via the ROS-Dependent Caspase Pathway
- Beyond the Fly: Fusome Conservation and Its Implications for Germ Cell Biology
- Muscle satellite cell dysfunction in critical illness: niche biology, regenerative failure, and implications for recovery
- Organoids and organs-on-a-chip as new approach methodologies at the convergence of cell biology and bioengineering
BibTeX
@article{rosen2026universal,
title = {{Universal cell embedding provides a foundation model for cell biology}},
author = {Yanay Rosen and Yusuf Roohani and Ayush Agrawal and Leon Samotorcan and Stephen R. Quake and Jure Leskovec},
journal = {Nature},
year = {2026},
month = jul,
doi = {10.1038/s41586-026-10689-z},
url = {https://doi.org/10.1038/s41586-026-10689-z},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Universal cell embedding provides a foundation model for cell biology
AU - Yanay Rosen
AU - Yusuf Roohani
AU - Ayush Agrawal
AU - Leon Samotorcan
AU - Stephen R. Quake
AU - Jure Leskovec
PY - 2026
DA - 2026/07/08
JO - Nature
DO - 10.1038/s41586-026-10689-z
UR - https://doi.org/10.1038/s41586-026-10689-z
AB - Developing a universal representation space for cells that encompasses the tremendous molecular diversity of cell types across species would be transformative for cell biology. Recent work using single-cell transcriptomic approaches to create molecular definitions of cell types in the form of cell atlases has provided the necessary data for such an endeavour1–3. Here we present the universal cell embedding (UCE) foundation model. UCE was trained on a large corpus of cell data using self-supervision, creating a unified biological latent space that can represent cells across diverse tissues and species. This latent space captures important biological variation despite the presence of experimental noise. UCE’s universality means that new cells can be embedded with no data labelling, model training or fine-tuning. We used UCE to create the Integrated Mega-scale Atlas, embedding 36 million cells, with more than 1,000 uniquely named cell types, from hundreds of experiments, dozens of tissues and eight species. We gain insights into the organization of cell types and tissues within the space. UCE’s embedding space exhibits emergent behaviour, identifying biology that it was never trained for, such as identifying developmental lineages and embedding data from species that were not included in the training set. Overall, by enabling a universal representation for every cell state and type, UCE is a valuable tool for analysis, annotation and hypothesis generation over single-cell data. The universal cell embedding foundation model learns to capture the organization and variation of cells by training on 36 million cells from hundreds of experiments, dozens of tissues and eight species.
SN - 0028-0836
SN - 1476-4687
ER -