This work stands somewhat apart: nearby in the archive sit studies of aging based on epigenetic marks, gene expression, and cellular composition, but no one had tried to quantify transcriptome damage itself. Neighboring papers use "aging clocks" built on DNA methylation or expression levels, yet those approaches capture correlates of age rather than causal mechanisms. This paper fills the gap: it offers a direct gauge of molecular damage that can be applied to RNA sequencing data that already exists.
The authors took publicly available RNA sequencing datasets from humans (16,020 samples from 26 tissues in the GTEx project) and mice (947 samples from 17 tissues in the Tabula Muris Senis consortium). For each sample they counted the fraction of reads supporting four types of damage: premature stop codons caused by intron retention, disruption of protein domains through abnormal splicing, expression of repetitive elements (LINE, LTR), and gene fusions. They then combined these measures into an overall damage index and trained machine learning models (LightGBM and elastic net) to predict chronological age from the damage profile.
All four types of damage correlate positively with age in most human and mouse tissues. Overall damage levels rise with age in nearly all tissues; the lowest levels appear in the pancreas, muscle, and blood, while the highest are in the cerebellum. In humans, sex differences turned up: women show higher damage in the brain, liver, muscle, and nerves, men in the esophagus. Cell composition influences damage, but even after adjusting for it the link to age holds, which means damage is an intrinsic feature of aging rather than a byproduct of shifting cell populations. The clocks predict age with a correlation of 0.82 (LightGBM) and 0.77 (elastic net) in mice. The clocks respond to interventions: in mice with accelerated aging (a BubR1 mutation) predicted age is elevated, while calorie restriction, rapamycin, and other anti-aging treatments lower it. During mouse embryonic development there is a temporary dip in predicted age (rejuvenation) between 10 and 16 days of gestation. In humans, a blood-based clock showed accelerated aging in patients with Alzheimer's disease compared with healthy people and those with mild cognitive impairment; conventional clocks based on gene expression revealed no such difference.
What sets this work apart from neighboring studies is that it measures not correlates of age but molecular damage itself: errors in RNA that directly threaten protein function. It is the first scalable quantitative assessment of transcriptome damage, applicable to any existing RNA sequencing data. This approach opens a path toward empirical measurement of biological entropy, a quantity long discussed but previously unmeasurable. What is more, the damage clock proved more sensitive to pathological aging (Alzheimer's disease) than traditional clocks based on gene expression.
The field could move toward combining different types of molecular damage: transcriptomic, proteomic, genomic. If protein oxidation and DNA mutations are added to RNA errors, one could build a "deleteriome," a complete map of age-related damage. Another direction is the search for specific regulators of RNA processing that account for the reversibility of damage under anti-aging interventions. Finally, the damage clock could be adapted to other species and tissues, including cell cultures and organoids.
The method is applicable in gerontology for assessing the effectiveness of anti-aging therapies, in neurology for early diagnosis of neurodegenerative diseases via blood tests, and in pharmacology for screening drugs that slow molecular aging.
The immediate steps include validating the damage clock on longitudinal data (the same people or animals at different points in time), as well as testing its predictive power for mortality and age-related diseases. Beyond that, it will be necessary to determine which RNA quality control mechanisms fail first and whether they can be restored.
The work speaks directly to a fundamental problem in gerontology: is the accumulation of molecular damage a cause or a consequence of aging? The proposed gauge makes it possible to test causal relationships in intervention experiments. The paper also touches on the question of biological age: how reversible is it, and which layers of molecular regulation can be rejuvenated?
The cerebellum turned out to be the most "damaged" region of the brain, even though the samples did not differ in age from other regions. Researchers link this to the near-total absence of neuron turnover in the cerebellum: the cells live a long time and accumulate errors.
The idea of measurable molecular damage as a clock of aging echoes Greg Egan's novel Permutation City, in which characters try to control biological age by manipulating cellular processes. In reality, though, we are still far from that kind of precision.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.