Overview
- Researchers at CeMM and the University of Vienna trained deep‑learning models on 25,712 high‑resolution histology images from 983 donors to create organ‑specific 'tissue clocks' that estimate an organ's biological age from microscopic architecture.
- The tissue clocks estimated age with a mean error of about 4.88 years and produced organ‑age gaps that matched shorter telomeres and known disease patterns, for example older pancreas in type‑2 diabetes and older brain in Alzheimer’s.
- Models were validated on independent tissue samples with variable accuracy across organs (correlations reported: brain 0.56, lung 0.76, skin 0.46), showing the method works better for some organs than others.
- The team built blood gene‑expression models to infer organ ageing and tested them on nine external datasets totaling 1,205 blood samples; these models show promise for aggregate or gut/spleen signals but cannot yet yield reliable per‑organ ages.
- Key limits remain: the analysis is cross‑sectional, many samples came from deceased donors with male overrepresentation, and blood predictions lack absolute calibration against matched living tissues, so longitudinal and better‑matched cohorts are needed before clinical use.