Home / CytoHub.AI
CytoHub.AI
AI that learns from living human tissue.
Every customer run of real human heart tissue adds to what CytoHub.AI learns.
- 3 AI tools live
- 4 AI patent families
- Next CytoRNA-FM
The feedback loop
Many readouts. One picture of the cell.
Readouts from the same tissue show how structure, function and genes move together. CytoHub.AI learns those links.
Measure Live
Up to five readouts from the same living human tissue, linked to compound, dose and batch.
Learn Live
CytoHub.AI places every cell on one map of human cell biology.
Predict Building
Cell states, disease signatures and drug response, before the next experiment runs.
Test Building
Predictions choose the next CardioVive experiment. Its results correct the model.
Live today
Three tools, already in use.
CytoMOTION
Contraction kinetics measured from video of beating tissue: rate, force and relaxation.
CytoVISION
Image-based phenotyping that scores structure and cell composition across every well.
CytoGPT
Answers questions across your study data and the scientific literature, with sources.
CytoRNA-FM · foundation model · building
A heart model on a proven foundation.
CytoRNA-FM builds on scVision, a single-cell foundation model from our Stanford collaborators, and is being tuned on CardioVive heart data that no public atlas contains.
- Cells as imagesEach cell is painted on a fixed map of 10,816 genes.
- Learned without labelsAbout 86M parameters, pre-trained on 72M public human cells.
- Disease that transfersA cardiomyopathy signature carried across disease subtypes.
scVision results in the preprint
Zero-shot, frozen encoder, public data. Heart-specific results follow fine-tuning.
Roadmap
From tools to a digital twin.
- LiveAI tools
CytoMOTION, CytoVISION and CytoGPT in use on partner studies.
- NextHeart-tuned CytoRNA-FM
The foundation model fine-tuned on CardioVive single-cell data.
- In developmentSafety digital twin
Simulate how human heart tissue responds to a compound before it is dosed.