TLPath predicts telomere length in human tissues from histopathology images
Telomere dysfunction is a key hallmark of aging linked to numerous age-related diseases including cardiovascular disorders, pulmonary fibrosis, and metabolic syndromes. Despite decades of research yielding strong evidence linking telomere biology to aging processes, the field faces a critical bottleneck: current telomere measurement methods require specialized molecular techniques that prevent large-scale studies and clinical implementation. Here we present TLPath, a novel deep learning framework that extracts normal tissue architecture from routine histopathology (H&E) images to predict bulk-tissue telomere length. Trained on the Genotype-Tissue Expression cohort comprising >7.3 million patch images from >5,000 whole-slide images across 919 individuals, TLPath makes a remarkable discovery: the extracted morphological features spontaneously separate young, middle-aged, and elderly individuals within most tissue types-- demonstrating for the first time that aging causes substantial architectural changes in tissues detectable without explicit age supervision. These extracted features can predict bulk-telomere length with significant accuracy (>0.51 in well-represented tissues), outperforming chronological age as a predictor (correlation = 0.20) and identifying age-discordant cases - detecting both accelerated telomeres shortening in young individuals and preserved telomeres in older individuals. Mechanistic interpretation reveals that TLPath leverages established senescence morphological markers, including nuclear-to-cytoplasmic ratio and nuclear shape variation, for its predictions. We applied TLPath in [~]2,800 new GTEx biopsies where concordant with known association, the predicted telomere length is shorter across most tissues from individuals with Type 1/2 diabetes. Overall, we demonstrate that aging substantially alters tissue morphology, which TLPath captures and uses to predict telomere length, enabling large-scale telomere biology studies using existing tissue archives.