Tabula — приватная foundation-модель для генетики клеток и поиска факторов омоложения

Researchers have presented Tabula — a foundation model for single-cell analysis that trains on data from different institutions without it ever leaving them, and with which they identified potential rejuvenation factors for human cells. The preprint was published on arXiv in July 2026.

What is the Tabula model

Tabula is a privacy-preserving foundation model for single-cell genomics that explicitly accounts for the tabular structure of single-cell data. The authors note that, unlike text, single-cell data is unordered and has a unique tabular form that existing models ignore. Tabula is built on federated learning — an approach in which several organizations jointly train one model without publishing raw data.

• Model — Tabula, a privacy-preserving foundation model for single cells
• Method — federated learning plus explicit modeling of the data's tabular structure
• Platform — Chiron, a decentralized system built on AI agents for collaborative training across institutions
• Data — a new scRNA-seq dataset of paired young and old human fibroblasts
• Biological systems — hematopoiesis, pancreatic endogenesis, neurogenesis, cardiogenesis

How to train without leaking data

Chiron — a decentralized platform built on AI agents that the authors developed specifically for Tabula — handles collaborative training without exchanging raw data. Institutions train a shared model on their private single-cell datasets without uploading them externally, removing the main barrier to scaling genomic models where patient data is protected. According to the preprint, beyond strong results on downstream benchmarks, Tabula uncovers combinatorial regulatory logic across four biological systems at once: hematopoiesis, pancreatic endogenesis, neurogenesis, and cardiogenesis.

What the model found about aging

Tabula proposed potential rejuvenation factors using a new scRNA-seq dataset of paired young and old human fibroblasts. The model ranked candidates in silico using age- and identity score-guided prioritization and, according to the authors, outperformed classical methods for this kind of selection. This is specifically about nominating candidates — genes and factors worth testing in experiments to revert old cells to a younger state — not a ready-made therapy.

"Tabula represents an important step in single-cell modeling,

combining tabular learning with federated learning and paving the way toward private virtual cells for human health," the abstract of the arXiv preprint states.

What this means

Tabula shows how to combine two hot topics — foundation models and data privacy — in biomedicine: organizations can train a shared model on private genomic data, and its output in the form of rejuvenation candidates revives interest in "virtual cells" as a tool for finding anti-aging therapies.

Frequently asked questions

How does Tabula differ from other single-cell models?

Tabula explicitly models the tabular structure of single-cell data, which, according to the authors, existing foundation models ignore, and it trains via federated learning without transferring raw data between institutions.

What are the rejuvenation factors that Tabula found?

These are candidate genes that the model selected from data on young and old human fibroblasts, using age- and identity score-guided prioritization. They still need to be tested experimentally — this is a prioritization for further research, not a ready-made therapy.