Transcriptomic noise accumulates within tissue identity across human aging: a systemic signature distinct from cell-composition drift

Aging is often described in two competing languages: accumulation of specific damaged or senescent cell populations versus systemic regulatory erosion affecting every cell. We provide a direct quantitative test of this dichotomy using variance-decomposition of bulk transcriptomes across age. In 263 GTEx v8 donors (20-79 years) with matched samples in six tissues, tissue identity accounts for approximately 0.73 of transcriptomic variance and declines by only 0.031 over forty years (variancePartition REML 0.789 to 0.758, PERMANOVA R-squared = 0.858; observed pi is 243-fold above a permutation null). The small decline is absorbed almost entirely by within-tissue, within-donor residual variance (pi_residual 0.168 to 0.194), not by between-donor systemic factors (pi_donor stable at approximately 0.064). The signature is therefore systemic noise accumulating within every cell type, not a shift toward outlier populations.

At the single-cell level, the residual growth is partly compositional (cell-type proportions shift) and partly cell-intrinsic: two platforms from the Tabula Muris Senis (Smart-seq2 and 10x Chromium) give complementary reads, with 10x -- which has lower age-related gene-detection bias -- showing Delta_pi approximately -0.07 per cell type in balanced analysis. Per-tissue, noise accumulates at very different rates: hematopoietic tissue grows in variance approximately 3-fold faster than skeletal muscle (Delta_var = +0.079 vs +0.028 over 40 years), but rates do not cleanly track a dividing/post-mitotic axis. A central unexpected result is that left ventricular myocardium shows the largest Delta_var of any tissue tested (+0.121) despite being post-mitotic, implicating non-turnover mechanisms -- age-related immune infiltration, fibroblast activation, and progressive fibrosis -- as the dominant drivers in cardiac aging at the transcriptomic-composition level.

Across four mammalian species (mouse, rat, macaque, human), the rate of pi_tissue decline scales inversely with maximum lifespan (alpha = -1.02 +/- 0.24, R-squared = 0.90, Spearman rho = -1.0). Caloric restriction in rat partially reverses aging-associated pi loss in marrow by reducing residual variance rather than restoring tissue-specific variance (87% rescue; bootstrap 95% CI 82-91%; mechanism confirmed in 100% of iterations). Our transcriptomic measurement captures tissue-specific noise accumulation rates that are not represented in methylation clocks, which are themselves largely tissue-invariant; the two signals are therefore complementary. We interpret the results as direct quantitative support for the systemic-noise view of aging, without denying the reality of specific senescent populations, and identify the hematopoietic compartment and left-ventricular myocardium as tissues worth prioritising for mechanistic follow-up.