Genomics & Single-Cell · Multi-institution consortium
CellFM
A 100-million-cell foundation model for pan-tissue single-cell annotation and cross-dataset integration.
What CellFM is
CellFM is a genomics foundation model trained on 100 million single cells to perform pan-tissue cell annotation and cross-dataset integration. Developed by a multi-institution consortium, it provides harmonised cell type labels across diverse studies.
Single-cell datasets frequently present integration challenges due to technical variations and heterogeneous sources across distinct tissues and experiments. While earlier models often focus primarily on human data, CellFM incorporates a broad corpus across tissues and studies, allowing it to annotate unseen cellular atlases.
How it works
The model is trained on a broad corpus of 100 million cells spanning diverse tissues and studies to enable generalised pan-tissue transfer and integration.
Key facts
- Version
- 1.0
- Domain
- Genomics
- Modality
- Single-cell gene expression
- Training data scale
- 100 million cells
- Required GPU memory
- 40 GB
What you can use it for
- Cross-dataset integration
- Integrate heterogeneous single-cell expression matrices across different studies into a unified representation.
- Pan-tissue cell type annotation
- Assign harmonised cell type identities to unlabelled or unseen single-cell tissue atlases.
Run it here
No installation and no GPU of your own: fill in the inputs and it runs on our servers. Results stay in your browser unless you sign in.
Not runnable yet
This model isn't available to run right now.Run this model
CellFM
Strengths and limits
Good at
- Trained on a 100-million-cell corpus spanning multiple tissues and experimental studies.
- Capable of annotating cellular atlases not present in its training data.
- Harmonises cell type labels across heterogeneous datasets.
Watch out for
- Requires substantial hardware resources, needing at least 40 GB of GPU memory.
- Input is restricted to expression matrices supplied in h5ad format.
Inputs and outputs
You provide
Expression matrix (h5ad URL)
You get back
Harmonised cell type labels across datasets.
Licence
Commercial use permitted · Apache-2.0
The Apache-2.0 licence allows use in commercial research and products.
Questions
Can I use CellFM commercially?
Commercial use permitted. The Apache-2.0 licence allows use in commercial research and products.
How do I run CellFM without installing anything?
Use the form under “Run it here” on this page: fill in the inputs and press Run model. It runs on our servers on 2× A100, typically in 8.0min, and a run costs about 2,800 credits. Visitors without an account get 1,000 credits a month.
What inputs does CellFM require?
It accepts an expression matrix provided via an h5ad file URL.
What outputs does CellFM produce?
It outputs tabular annotations containing harmonised cell type labels across datasets.
How does CellFM compare to models like scGPT and Geneformer?
Unlike human-centric models such as scGPT and Geneformer, CellFM is trained across a wider diversity of tissues and studies to support unseen atlas annotation and cross-dataset integration.
What hardware is necessary to run CellFM?
Execution requires a GPU with at least 40 GB of memory.
Written from CellFM's own documentation. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.