Brian Reed: Hey — what did you make of that Insilico paper?
Eliza Ward: Still processing it, honestly. The core result is — six AI models estimated patients' biological age from blood proteins, and after rentosertib, all six said the treated patients looked younger. Published September 7th in Nature Biotechnology.
Brian Reed: All six. Same direction.
Eliza Ward: Some clocks put it at three to four years younger. One clock said up to six. And look, the strongest signal wasn't even at the end — it peaked around four weeks, on the 30 milligram twice-daily dose.
Brian Reed: Peaked at four weeks and then what — came back down?
Eliza Ward: That's not fully resolved in the paper. But here's the grounding fact: this is 42 people. Subgroup from a 71-patient lung disease trial. Idiopathic pulmonary fibrosis.
Brian Reed: Right — so not an aging trial. A drug for a progressive lung condition that incidentally rang every aging clock they measured.
Eliza Ward: Which is either the most interesting finding in a long time, or — it's all six trackers sharing the same blind spot.
Brian Reed: And that shared blind spot is — let me see if I can pin this — if all six clocks are reading fibrosis-related proteins, inflammation markers, senescence signals, then they're not six independent confirmations of an aging effect. They're one signal read through six lenses.
Eliza Ward: Which is exactly what nobody's tested. Strip out the fibrosis proteins, run the clocks again — does the aging signal hold?
Brian Reed: Right — and that's the missing experiment. But here's what is actually new, and I don't want to lose it in the skepticism: rentosertib isn't rapamycin. It isn't metformin. Those are existing drugs getting repurposed into longevity trials. This molecule — TNIK as the target, the whole thing — was built from scratch by Insilico Medicine's AI. De novo. That's genuinely the first time something fully AI-native has reached clinical evaluation.
Eliza Ward: Wait — and TNIK sits at the intersection of fibrosis and six hallmarks of aging. That's why the dual signal isn't a coincidence. The target was chosen partly because of that overlap.
Brian Reed: Which makes it harder to separate. The aging signal might be baked into the target selection, not discovered in the patients.
Eliza Ward: And the Nature Medicine paper from June 2025 — that already confirmed the lung function piece. Dose-dependent Forced Vital Capacity improvement, 71 IPF patients, safety endpoint met. So the aging clocks in Nature Biotechnology are layered on top of a disease result. The question is whether they're telling us something additional or just — reflecting the same underlying improvement.
Brian Reed: The part I don't get — and this feels important — these clocks aren't regulatory endpoints. The FDA doesn't recognize them as proof of anything. They're pharmacodynamic biomarkers. Exploratory ones. So what exactly did Nature Biotechnology publish?
Eliza Ward: An interesting biological signal in 42 people that the FDA has no current framework to act on. Because aging isn't an approvable indication.
Brian Reed: So the Phase 3 — 320 people, 52 weeks, measuring lung function — that trial isn't designed to validate the aging signal. It's an IPF trial. Which means even if the clocks replicate in a bigger population, it doesn't move a regulatory needle.
Eliza Ward: Not just that — the take already circulating is that six clocks agreeing is basically proof. That rentosertib slows aging in humans. And I want to push back on that hard, because it doesn't hold.
Brian Reed: Walk me through it concretely.
Eliza Ward: Okay — picture a 67-year-old IPF patient, clinical center in China, her lung scarring is slowing on the drug. Her blood proteins shift. Six aging clocks read her as biologically younger. But you fundamentally — wait, here's what you cannot tell: is her aging biology reversing, or are her tissues just no longer being actively destroyed by fibrosis? Those are different things.
Brian Reed: Right — and IPF onset is typically around 65. These are patients already in fibrotic crisis. So the baseline biology is catastrophic.
Eliza Ward: Exactly. And rentosertib has never been tested in healthy, non-IPF populations. So the aging signal cannot currently be disentangled from disease treatment. That's not a nitpick — that's the whole question.
Brian Reed: The part I'd push back on — TNIK sits at the intersection of fibrosis and six hallmarks of aging. Dual mechanism is at least biologically plausible, not just coincidental noise.
Eliza Ward: Plausible, yes. Separable — no. Nobody has actually stripped the fibrosis-related proteins from those clock models and checked whether the aging signal survives. That experiment hasn't been run. Plausibility isn't separability.
Brian Reed: And then there's Zhavoronkov's number. Three added years of life expectancy, scaled globally — he says that's 25 billion additional life years. That's not epidemiology. That's founder arithmetic.
Eliza Ward: It's arithmetic that lands right next to a Nature Biotechnology paper, which is — I mean, that proximity does a lot of work commercially. The actual Phase 3 question — what a 320-person IPF trial can even tell us about aging biology — that's the part we need to get into.
Brian Reed: And that's the concrete mismatch, right — Insilico compressed the discovery piece to 18 months, target to candidate, genuinely faster than anything conventional. But Phase I, Phase IIa, Phase 3 — those timelines didn't move. The clinical bottleneck is completely unchanged. So the AI acceleration happened entirely in a part of the pipeline that was never the slowest part.
Eliza Ward: Wait — that's actually the thing people keep glossing over.
Brian Reed: And then the Phase 3 — 320 participants, 52 weeks, the primary endpoint is lung function in IPF patients. That's what it's powered to detect. Not biological age reduction. So even if a researcher is sitting in, I don't know, a monitoring committee meeting in late 2027, watching clock data come in on those 320 patients, she literally cannot call it aging validation. The trial wasn't designed for that question.
Eliza Ward: No aging indication pathway at the FDA means — the geroprotective data has nowhere to go regulatorily. It's commercially loud, I mean the Nature Biotechnology paper is doing a lot of work for Insilico's valuation, but it's regulatorily mute.
Brian Reed: So rentosertib's only real near-term path is IPF approval. That's it.
Eliza Ward: Right — and here's where the Nature Biotechnology authors actually do something interesting. They're not claiming approval. They're proposing that proteomic aging clocks get embedded in standard disease trials as a methodological framework. Bolt the clocks onto an ongoing trial, collect the signal systematically, build the evidence base. It's a proof-of-concept architecture, not a regulatory filing.
Brian Reed: Which is — okay, that's genuinely useful science. But it doesn't resolve the FDA problem. That's a methodology contribution, not a pathway.
Eliza Ward: No. If the FDA framework ever shifts to accommodate aging endpoints, rentosertib's signal could be a template. If it doesn't — and there's no current sign it will — the aging findings stay exploratory regardless of how many trials replicate them.
Brian Reed: So the concrete thing to watch isn't Phase 3 aging data. It's whether the FDA opens any kind of aging indication track at all — because without that, the most interesting biological signal Insilico has found doesn't have a door to walk through.
Eliza Ward: And nobody in this story controls that. Not Insilico Medicine, not Zhavoronkov, not the Nature Biotechnology authors. The FDA framework moves — or it doesn't — on its own timeline, for its own reasons, based on evidence that doesn't exist yet.
Brian Reed: Which is — I mean, that's a strange place for a company to be. You've run the fastest de novo AI drug discovery on record, 18 months target to candidate, you've got a signal in six aging clocks, you're published twice in Nature journals, and the thing that determines whether any of that matters is a regulatory body that currently has no mechanism to evaluate it.
Eliza Ward: That's — yeah. The science is ahead of the framework. And the 42-person signal in that September 7th paper, ambiguous as it is, actually becomes the most important data point in a new category if the FDA ever opens an aging indication track. Same numbers. Completely different status.
Brian Reed: And if it doesn't move — rentosertib is an IPF drug that produced a fascinating footnote. That's actually where we are right now.
