Organs age at different rates

The work is based on so‑called “tissue clocks”: models based on artificial intelligence that estimate the biological age of a specific organ based on the microscopic structure of the tissue. The researchers analyzed more than 25,000 images of tissue sections covering 40 different types: from the brain and heart to the lungs, pancreas, skin, and intestines. The material used was the collection from the Genotype‑Tissue Expression (GTEx) project, which includes samples from 983 individuals. In total, the AI processed approximately 480 million individual image fragments using state‑of‑the‑art computer vision models.

Age turned out to be the most significant factor determining the appearance of tissue, and this effect was observed in all 40 tissue types, even without special training of the model on age markers. Based on this, the team created “tissue clocks” capable of predicting the biological age of an organ based on its histological pattern. The average prediction error was only 4.9 years. At the same time, such clocks are more accurate than estimates based on DNA methylation; they reflect tissue‑specific pathology and correlate with known signs of aging — telomere shortening, the presence of pathologies in tissues, and the number of chronic diseases in a person.

The analysis showed that different organs age according to different scenarios. For example, the lungs, kidneys, pancreas, and adrenal glands show signs of accelerated aging already between the ages of 20 and 40. In the uterus, a particularly noticeable shift occurs around the time of menopause. In some tissues, the peaks of accelerated aging appear later in life. Furthermore, researchers have identified clear links between tissue‑specific aging and diseases: kidney failure was associated with accelerated aging in several tissues at once, and diabetes was associated with pronounced changes in the pancreas.

An important step was that it will now be possible to do without invasive tissue sampling. The gene expression profiles in blood were compared with the biological age of the same individuals, calculated from tissue images, and predictors were created that make it possible to determine tissue‑specific biological age based solely on a blood sample. According to the researchers, this is a conceptual breakthrough, as the language of tissue aging has been translated into a format that is accessible during a routine blood draw.

Such blood markers successfully identified aging patterns associated with a number of diseases. Thus, in Alzheimer’s disease, the strongest signal of aging was detected precisely in the brain, while in Crohn’s disease it was detected in the gastrointestinal tract. This shows that the approach not only allows for assessing the overall rate of aging but also enables its localization in specific body systems, linking it to specific pathologies.

The authors argue that aging is a complex process in which systemic and tissue‑specific factors act differently in each organ. The architecture of the tissue integrates numerous molecular and physiological changes associated with age and disease. In the future, such methods may form the basis for minimally invasive diagnostics: using a routine blood test, doctors will be able to monitor the condition of individual organs and the progression of diseases.

The work also demonstrates the growing potential of artificial intelligence in pathology and aging research. Combining large datasets, such as histological images, gene expression profiles, and clinical data, makes it possible to obtain a comprehensive picture of how aging manifests itself in different parts of the human body.