AIBRIEF
An AI trained on 25,712 tissue images from 983 donors reads each organ's biological age from a single blood draw, with a five year error band and no clinical test yet.
Your body does not have a single age. A Nature Medicine study published this week, led by André Rendeiro with co-first authors Ernesto Abila, Iva Buljan, and Yimin Zheng, trained AI "tissue clocks" on 25,712 microscopic images from 983 deceased donors across 29 organs. The clocks estimate whether each organ looks biologically older or younger than the donor's chronological age. On average, the estimates run about five years off.
The model captures aging patterns that the calendar cannot. Tissues flagged as older showed shrinkage, scar-like stiffening, loss of tiny blood vessels, fat accumulation in skeletal muscle, and thinning nerves. A 60-year-old might carry a lung that looks 65, while the rest of the body holds closer to its birthday. Some donors aged uniformly; others showed a single organ drifting far ahead of the rest, the paper reports.
To skip a biopsy, the team matched each tissue-clock reading to gene-activity patterns in the donor's blood, building a single-draw proxy. The university release and ScienceAlert's coverage translate that into "one blood test for organ age" headlines. The method is not that, yet: News-Medical notes the readout is a research proxy, not a hospital assay, and cannot track an organ's aging rate over time. The model and code are released on Zenodo as tissue-clocks v1.0.0 for independent follow-up.