New AI Model TLPath Predicts Biological Aging by Measuring Telomeres From Standard Medical Biopsy Scans

New computational model TLPath measures telomere length from tissue images. Discover how AI identifies cellular aging signatures faster and cheaper than lab tests.

By: AXL Media

Published: Mar 16, 2026, 12:16 PM EDT

Source: Information for this report was sourced from Sanford Burnham Prebys

Decoding the Physical Architecture of Cellular Aging

The quest to quantify biological aging has traditionally required expensive and specialized laboratory procedures to measure telomeres, the protective caps at the ends of our chromosomes. However, a breakthrough study published in Cell Reports Methods reveals that the secrets of aging are hidden in plain sight within the structure of our cells. Researchers at Sanford Burnham Prebys have introduced TLPath, a computational model that infers telomere length by analyzing high-resolution images of routine medical biopsies. This innovation rests on the discovery that as telomeres shorten, they leave behind "architectural fingerprints"—measurable changes in tissue and cell shape that can be identified through advanced computer vision.

The Biological Function of Telomeres as Genetic Buffers

Telomeres serve as essential bumpers that prevent the loss of critical genetic information during cell division. As Dr. Sanju Sinha explains, every time DNA replicates, a small portion at the end of the strand cannot be copied. To protect the genome, cells evolved repeating DNA sequences at the terminal ends that can be "whittled down" instead of functional genes. Over a lifetime, the gradual shortening of these buffers correlates with a person's risk for chronic age-related diseases. By accurately predicting this length, TLPath allows researchers to assess a patient's biological age—the actual wear and tear on their body—rather than just their chronological age in years.

Training AI on Massive Datasets of Human Tissue

To build TLPath, the research team utilized the Genotype-Tissue Expression Project, a massive NIH initiative. They trained the model on 5,263 histopathology slides representing 18 different tissue types from nearly 1,000 individuals. The AI segments each high-resolution scan into an average of 1,387 small patches, scouring each one for up to 1,024 distinct structural features. By weighting these features, the model learns to associate specific tissue patterns with the paired laboratory data of telomere length. This "foundation model" approach allows the AI to recognize complex, high-order features that are often invisible to the human eye but highly predictive of genetic health.