Panoramic modelling of single-cell transcriptomes at scale with CellNet

Abstract

Characterizing cellular diversity holistically is critical for understanding tissue development, identifying disease-associated cell populations, and investigating gene programmes linked to specialized functions. Recent advances in single-cell profiling have created unprecedented opportunities to uncover this diversity by providing a comprehensive view of the transcriptome. However, existing foundation models generally leverage only a restricted portion of this increasingly rich molecular information for cellular representation, as a trade-off for computational efficiency. Here we present CellNet, which enables panoramic modelling of single-cell transcriptomes by integrating DNA-derived gene identities and measured RNA expression into a shared image representation. Pretrained on 231 million image-based cell profiles from 40 species, CellNet improves the recognition of rare populations and condition-associated states in evaluated cohorts while retaining strong general cell representations. It predicts withheld coding-gene expression and selected non-coding targets in total-RNA data, and recovers lineage-specific expression patterns in spatial measurements. Prediction also extends across species, including to targets without expression exposure in pretraining. Its gene representations retrieve known non-coding modules and developmental or neural programmes, even from sparsely detected RNA queries. CellNet provides a foundation for cell-biology discoveries at the whole-transcriptome scale through its panoramic modelling of single-cell transcriptomes. The CellNet website is available at https://www.cellnet.top.

Citation

Xiaoqing Lian, Pei Liu, Jiahui Tan, Luoda Tan, Chaoyi Li, Quan Zou, Chen Lin, Xiangxiang Zeng. Panoramic modelling of single-cell transcriptomes at scale with CellNet. arXiv:2610.21943 (2026). doi:10.48550/arXiv.2610.21943.

@misc{lian2026cellnet,
  title = {Panoramic modelling of single-cell transcriptomes at scale with CellNet},
  author = {Lian, Xiaoqing and Liu, Pei and Tan, Jiahui and Tan, Luoda and Li, Chaoyi and Zou, Quan and Lin, Chen and Zeng, Xiangxiang},
  year = {2026},
  eprint = {2610.21943},
  archivePrefix = {arXiv},
  primaryClass = {q-bio.GN},
  doi = {10.48550/arXiv.2610.21943},
  url = {https://arxiv.org/abs/2610.21943}
}