In a breakthrough that harnesses the power of artificial intelligence to unlock secrets held within our cells, researchers at the Sanford Burnham Prebys Medical Discovery Institute have developed a groundbreaking computational tool capable of predicting telomere length simply by analyzing tissue morphology. Published in the influential journal Cell Reports Methods in March 2026, this pioneering approach utilizes complex machine learning algorithms applied to high-resolution images of routine biopsy samples, presenting a new horizon in understanding the intricate relationship between cellular structure and the aging process.
Telomeres, the repetitive DNA sequences capping the ends of chromosomes, function as crucial protective buffers preventing the loss of genetic information during cell division. Each time a cell replicates, telomeres shorten slightly, which has long been associated with cellular aging and the onset of various age-related diseases. Until now, measuring telomere length has involved intricate, costly laboratory tests that are difficult to perform at scale. The novel tool, named TLPath, circumvents these limitations by linking morphological changes observable in tissue slides to telomere length, using routinely generated histopathology data.
At the heart of TLPath’s innovation is its ability to analyze more than 5,000 whole-slide images from 919 individuals spanning 18 different tissue types. By dissecting each slide into over a thousand smaller patches, the model extracts upwards of a thousand structural features from each area. These features represent complex tissue architectures, subtle cellular arrangements, and microscopic structural nuances that human pathologists may not readily discern but are meaningful for computational detection. Through statistical weighting of these features, TLPath develops a robust predictive score correlating tissue morphology with telomere length.
Leveraging the extensive dataset from the Genotype-Tissue Expression Project, an NIH initiative designed to elucidate genetic influences on disease, the researchers trained TLPath on this rich repository of paired histopathological images and telomere length measurements. This integration enabled the model not only to learn characteristic morphologies associated with varying telomere lengths but also to generalize its predictive power across tissue samples not included in the initial training, demonstrating remarkable accuracy and robustness.
One of the most striking revelations from this research is that TLPath outperforms chronological age as a predictor of telomere length. This finding underscores the nuanced complexity in biological aging that chronological time alone cannot capture. Importantly, the tool successfully distinguishes telomere length differences among individuals of identical ages, pointing to the model’s sensitivity in capturing biological aging markers beyond conventional demographic data.
Behind this technological leap lies recent advances in computer vision and foundation models in histopathology. Unlike traditional image analysis focusing on individual pixels, foundation models abstract high-level, composite features, capturing interrelations and patterns within tissue architecture. These newly discerned features, although only partially interpretable by humans at present, hold validated predictive relevance, enabling TLPath to mine meaningful biological insights from the complex tapestry of cellular morphology.
The implications of TLPath extend far beyond telomere measurement. Since digitized histopathology slides are a mainstay of clinical diagnostics, the model’s success highlights the vast untapped potential in existing medical imaging archives worldwide. However, the current obstacle remains the widespread digitization and accessibility of these slides for research. If biobanks and clinical institutions can standardize scanning and sharing practices, TLPath could be deployed for large-scale biological aging studies and potentially integrated into clinical workflows to inform patient care.
Furthermore, this technology promises to democratize telomere research by drastically reducing the barriers to assessing telomere dynamics. Where conventional molecular assays require specialized equipment and expertise, TLPath offers a scalable, cost-effective alternative enabling population-scale investigations into the links between telomere attrition, disease susceptibility, and aging biology.
According to Dr. Sanju Sinha, an assistant professor leading this initiative, understanding telomere biology is pivotal to uncovering mechanisms of aging and age-associated diseases. His team’s work demonstrates that the physical manifestation of molecular changes within tissues — captured visually and quantified computationally — can serve as a surrogate marker for underlying genomic alterations. This conceptual shift from biochemical assays to morphological prediction charts a new direction in biomedical research.
The study was meticulously conducted by Anamika Yadav and Kyle Alvarez, who share first authorship, along with contributions from researchers Akanimoh Adeleye, Yu Xin Wang, and Michael Jackson. Their collective efforts were supported by the National Cancer Institute-designated Cancer Center at Sanford Burnham Prebys and the institution’s Lab Experience As Pathway to Graduate School Program.
Looking forward, the scientists envision extending TLPath’s capabilities to other genomic features and disease markers observable through tissue morphology. As artificial intelligence continues to reshape biomedical sciences, tools like TLPath exemplify how integrating advanced computational models with vast biological data sets can accelerate discovery, transforming how we approach aging and precision medicine in the years to come.
Subject of Research: Human tissue samples
Article Title: Tissue Morphology Predicts Telomere Shortening in Human Tissues
News Publication Date: 16-Mar-2026
Web References: http://dx.doi.org/10.1016/j.crmeth.2026.101336
References: DOI 10.1016/j.crmeth.2026.101336
Image Credits: Anamika Yadav, Kyle Alvarez, Sanju Sinha, Sanford Burnham Prebys
Keywords: Telomeres, DNA, Chromosomes, Medical tests, Biopsies, Pathology, Artificial intelligence, Deep learning, Machine learning
