What the Information Theory of Aging really claims — and what it doesn't

The Information Theory of Aging (ITOA) is one of the most quoted, and one of the most misquoted, ideas in modern biology. The press version says: aging is a software bug, the bug can be patched, and David Sinclair's lab has done it in mice. The actual claim, written down carefully in Lu et al. and elaborated in a half-dozen follow-up papers, is narrower and considerably more interesting. This essay walks through what the theory actually proposes, what the 2020 paper actually demonstrated, and where reasonable people in the field still disagree.

The claim, in one paragraph

ITOA proposes that the proximate cause of aging at the cellular level is the gradual loss of epigenetic information — specifically, the patterns of DNA methylation and chromatin organisation that tell each of your cells which of its ~20,000 genes to read and how loudly. Genetic information (the DNA sequence) is largely preserved over a human lifetime; epigenetic information is not. ITOA's strong form says that if you could restore the original epigenetic pattern in an old cell without erasing its identity, you would restore most of its youthful function. The Lu paper is a proof-of-concept that, in mouse retinal ganglion cells damaged by glaucoma, partial transient expression of three of the four Yamanaka factors (OCT4, SOX2, KLF4 — "OSK") does exactly that.

What the 2020 paper actually showed

Lu et al. delivered an inducible OSK cassette by AAV2 to the eyes of aged mice. Treated retinas, but not controls, regained axon-regeneration capacity after optic-nerve crush, recovered visual acuity in a glaucoma model, and showed a partial reset of the Horvath-style epigenetic age clock. Critically, the cells did not dedifferentiate: they remained retinal ganglion cells. They did not turn into stem cells, they did not form teratomas, and they did not lose retinal-specific gene expression.

Three details matter for anyone who wants to think clearly about this paper. First, the effect required transient expression: continuous OSK was either ineffective or actively harmful. Second, the inclusion of the fourth Yamanaka factor, MYC, abolished the rejuvenation effect and caused tumours. Third, the rejuvenation depended on TET-mediated active demethylation; if you blocked TET, you blocked the effect. These three facts together suggest that the cell already knows what its youthful methylation pattern was, and OSK simply gives it permission to read that record.

This is the most important sentence in the paper, and it is in the supplementary materials: 'The youthful methylation landscape is recoverable but not reconstructable; this argues for the existence of a stable backup.'

— Lu et al., Supplementary Discussion, 2020

What the paper did not show

It did not show lifespan extension. It did not show systemic rejuvenation of an old animal. It did not show that the same trick works in any tissue other than the retina. It did not show that a small molecule could substitute for the AAV-OSK cassette. And it did not show that any of this is safe to attempt in humans, where the regulatory framework around partial reprogramming remains, charitably, undefined.

Several follow-up studies — in particular Yang et al. (2023) in Cell — have extended the claim to other tissues and to the whole-organism level in mice, with cautious but real evidence of physiological rejuvenation. None of these studies has yet been independently replicated by a lab with no Sinclair-affiliated authors. That replication is the single most important experiment in the field for the next 24 months. Any reader who comes across a press release announcing OSK-based rejuvenation in humans before that replication exists should treat it as marketing, not science.

Why this matters for protocol design

Most readers of Vitaei do not have access to AAV-OSK. They have access to small molecules: NAD+ precursors, sirtuin activators, mTOR-modulators, senolytics. ITOA does not directly endorse any of these. It does, however, suggest a useful design heuristic. If aging is at heart a problem of epigenetic noise, then interventions that reduce demand on the epigenetic-maintenance machinery — caloric restriction, exercise, anything that quiets the chronic stress response — should bend the trajectory more than interventions that target a single hallmark in isolation.

This is why the protocols we recommend foreground sleep, exercise and dietary structure before they recommend any compound. The compounds matter, but the substrate matters more. A reader who optimises rapamycin dosing while sleeping six hours a night is fixing a symptom while burning the cause.

Where reasonable people still disagree

• Whether the Horvath methylation clocks measure aging itself or simply correlate with it. (Vitaei's editorial position: the clocks correlate, they do not yet diagnose.)
• Whether the 'backup copy' Sinclair invokes is real biology or a useful metaphor. (Position: there is suggestive evidence — heterochromatin organisation, nucleolar architecture — but no smoking gun.)
• Whether partial reprogramming is the right therapeutic vehicle, or whether direct epigenetic editing (CRISPR-dCas9 + DNMT/TET fusions) will eventually displace it. (Position: the editing approach is more elegant in principle but a decade further from clinical use.)
• Whether ITOA explains aging or only describes it. (Position: this is the single deepest open question in the field.)

How we score the underlying evidence

Vitaei rates ITOA as Tier II evidence: convincing in vivo demonstrations in mice, mechanistically coherent, but lacking human data and lacking independent replication of the most ambitious whole-organism claims. We rate the related rapamycin and NAD+ literatures as stronger (Tier I/II depending on endpoint) because the human exposure data is broader. We do not yet rate any commercial 'epigenetic reprogramming' product because no such product exists outside grey-market channels we will not link to.

Further reading

If you read one paper, read Lu et al. 2020. If you read two, add López-Otín et al. 2023, the consensus update of the hallmarks framework that situates ITOA among its peers. If you read three, add Yang et al. 2023, which extends the claim systemically and is currently the field's most cited follow-up. If you read four, add Mannick et al. 2014 on rapamycin in aged human immune systems — a useful corrective in a field that often forgets that controlled human data already exists for some of the major claims.