Single-Cell and Cell Imaging Foundation Models
Models pretrained on gene expression of single cells or on cell microscopy images, used to annotate cells and predict perturbations.
Tier Grades
Tier A: Frontier AI labs
feasible
Google with Yale fine-tuned Gemma 2 on over 57 million cells.
Tier B: Indonesian AI labs
stretch
Open checkpoints exist, but no Indonesian single-cell programme is observed.
Tier C: Institution units such as ITB
feasible
Running open Geneformer or C2S-Scale 2B on public atlases fits a university biology unit.
Tier D: Organisations such as SEMAI (PoT, PoC, PoV)
not feasible
Single-cell analysis is a research activity without a delivery use case in the list.
Where this comes from
Tiers: a) frontier AI labs, b) Indonesian AI labs, c) institution units such as ITB, d) organisations such as SEMAI delivering proof of technology, concept and value. Each rationale cites the anchors below.
Anchors
- ComputeC2S-Scale: fine-tuned from Gemma 2 at 2B and 27B parameters; training compute not stated.
- DataGeneformer V1: about 30 million single-cell transcriptomes, expanded model about 104 million; C2S-Scale: over 57 million human and mouse cells from over 800 datasets.
- Team size (proxy)Proxy only (author count of an exemplar paper, not a headcount): Not in the sources read (model cards give no author count).
- Time horizonYears.
- Shelf-life signalsC2S-Scale converts expression into text for a general LLM, an alternative to purpose-built models such as Geneformer.
- Pipeline roleTier A releases cell models; Tier C applies them to local datasets; Tier B and Tier D have no observed role.
Where this comes from
Team size is a proxy (author counts of exemplar papers), not a headcount.
Who and What
Exemplars
Observed actors
Linked models
Links
R&D TopicSingle-Cell and Cell Imaging Foundation Models
ModelsGeneformerscientific
C2S-Scale (Cell2Sentence)Yale University, scientific
VISTA-2D and VISTA-3DNVIDIA, visionC2S-Scale-Gemma-2-2BYale University, scientific+2 more
C2S-Scale (Cell2Sentence)Yale University, scientificVISTA-2D and VISTA-3DNVIDIA, visionC2S-Scale-Gemma-2-2BYale University, scientific+2 moreUse CasesMedical Image SegmentationHealthcare
Who Runs Itnone recorded
Models 6
C2S-Scale (Cell2Sentence)Yale University, Google, Google DeepMind, scientificexemplarC2S-Scale-Gemma-2-2BYale University, Google, Google DeepMind, scientificproduced byC2S-Scale-Gemma-2-27BYale University, Google, Google DeepMind, scientificproduced byVISTA-2D and VISTA-3DNVIDIA, visionproduced by
OpenPhenom-S/16Recursion, scientificproduced byGeneformerscientificexemplarPapers
- Cell Detection with Star-convex PolygonsUwe Schmidt et al., MICCAI 2018 (per arXiv comment), 2018 founding
Where this comes from
The papers to read first: where the direction began, its current best, and a survey when one exists.
Sources
- huggingface.co/ctheodoris/Geneformer/raw/main/README.md https://huggingface.co/ctheodoris/Geneformer/raw/main/README.md
- huggingface.co/vandijklab/C2S-Scale-Gemma-2-27B/raw/main/REA... https://huggingface.co/vandijklab/C2S-Scale-Gemma-2-27B/raw/main/README.md
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Single-Cell and Cell Imaging Foundation Models", https://aurion.semai.id/topics/single-cell-and-cell-imaging-foundation-models/, records of October 2026.