Inferring organ aging, hallmark of aging and senescence scores from blood and facial photographs

Abstract

Aging progresses asynchronously across organs, motivating the development of organ-resolved aging clocks. Direct assessment of organ aging in living humans, however, is largely impractical. Here, we construct tissue-specific transcriptomic aging clocks and infer organ biological age from paired whole-blood (blood-organ age) and tissue RNA sequencing data in the GTEx project. We observe high variability both in the strength of aging signatures across organs and in the ability to predict organ age and organ scores from blood. Pathway analysis shows that organ aging clocks are dominated by biological processes that differ across tissues. Age-associated multi-organ modules identified in an aging multi-organ network correlation analysis and organ age features causally mediated through organ-blood-organ transcriptomic interactions support a role for systemic signaling in coupling organ aging processes. Given the established role of cellular senescence as a driver of tissue dysfunction and its tissue-specific accumulation during aging, we extend this framework to estimate burden of organ-level senescence and hallmarks of aging. Using senescence- and hallmark-associated gene sets, we derive tissue-specific senescence and hallmark scores (organ scores) and demonstrate that these burdens across multiple organs can be inferred from blood transcriptomes (blood-organ scores). Finally, in pursuit of extending these models to other blood omics and alternative non-invasive assessments, we evaluate whether blood methylation (methylation-blood-organ age and scores) and facial photographs (facial-blood-organ age and scores) can serve as proxies for blood-organ age and blood-organ scores in independent cohorts (Edifice Health and Health and Retirement Study). The methylation- and facial-blood-organ age and scores were associated with mortality in additional testing cohorts (Framingham Heart Study and IMDB-WIKI dataset). These results establish a cross-modal framework for inferring organ-specific aging and senescence from minimally invasive data.