scFoundation
100M-parameter foundation model for single-cell transcriptomics, trained on 50M+ cells.
About
scFoundation is a large-scale foundation model for single-cell RNA sequencing data, published in Nature Methods in 2024. It packs roughly 100 million parameters covering about 20,000 genes, pretrained on more than 50 million human single-cell expression profiles gathered from GEO, the Human Cell Atlas, Single Cell Portal and EMBL-EBI. Its backbone, xTrimoGene, uses an asymmetric transformer with a read-depth-aware masking objective so it stays robust to the wildly varying sequencing depths across labs and protocols.
The model is used as a plug-in representation: frozen or lightly adapted embeddings feed task-specific models for gene-expression enhancement, drug-response prediction, perturbation forecasting and cell-type annotation, where the authors report state-of-the-art results. Code and models are published on GitHub, though one independent review notes the public repo does not fully reproduce the original pretraining run — treat the released weights as a starting point, not a bit-exact replica of the paper's model.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
- Paper-reported (Nature Methods, doi 10.1038/s41592-024-02305-7): state-of-the-art on gene-expression enhancement, tissue and single-cell drug response, perturbation prediction and cell-type annotation
- No independent standardized benchmark suite published
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (TIER metrics, metrics-v1). Evidence: peer-reviewed Nature Methods paper (doi 10.1038/s41592-024-02305-7) with six downstream task families evaluated; public code + model repo (biomap-research/scFoundation); widely cited in single-cell foundation-model literature. Strengths: peer review, large-scale pretraining, broad downstream coverage. Docked for: (1) an independent review flags that the pretraining run is not fully reproducible from the public repo; (2) all headline numbers are author-reported. License: not confirmed in sources checked (see repo).
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 5.0/10
- Openness · license + docs
- 6.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Nature Methods 2024 (doi 10.1038/s41592-024-02305-7); 100M params, ~20K genes, 50M+ cells
- · Public repo biomap-research/scFoundation; independent review notes pretraining not fully reproducible from public repo
- · Widely cited in single-cell foundation-model literature