Finn Brooks: Juniper, hey — I sent you that study at like 11pm Tuesday and I have zero regrets about that.
Juniper Vale: You did, and I read it at midnight like a person with no self-control, so.
Finn Brooks: Worth it though — because the thing that sent me is this: a team at UH Mānoa, led by Alika K. Maunakea at the John A. Burns School of Medicine, published in Scientific Reports on July 14th, and they are saying gut bacteria don't just correlate with aging — they track with the *pace* of aging. How fast the clock moves. And they have a number.
Juniper Vale: Fifteen percent.
Finn Brooks: Fifteen percent of the variation in DunedinPACE scores — which is the specific biomarker that measures aging rate, not static age — explained by gut microbial patterns. In 123 people. And I keep going back and forth on whether 15% is huge or whether I'm getting played by a small number.
Juniper Vale: You're not getting played, but you might be slightly over-rotating — and the EpiBiome models catching a signal only on DunedinPACE and not on the older clocks is the part I actually can't stop thinking about.
Finn Brooks: Right — because that means the microbiome is linked to aging *speed*, not just some fixed biological snapshot. That distinction is kind of wild.
Juniper Vale: It is. And that's what we're here to pull apart — whether this is the piece that changes how we think about aging, or one piece of a much bigger puzzle we haven't mapped yet.
Finn Brooks: But wait — that 'only DunedinPACE responded' thing, I keep almost letting that slide and I shouldn't. Like why didn't Horvath or Levine or GrimAge2 catch anything? They've been the workhorses of the whole field.
Juniper Vale: Think of it like this — the older clocks are a photograph. They tell you how old your cells look right now. DunedinPACE is a speedometer. It tells you how fast you're aging at this moment. And the microbiome, it turns out, is not a great predictor of a photograph — but it is connected to the speed.
Finn Brooks: Oh. Ohhhh. That's actually — yeah, okay, I get it now.
Juniper Vale: And EpiBiome — the machine-learning models they built — found zero predictive signal for Horvath, Levine, GrimAge2. Flat. Only DunedinPACE responded. Which, I mean, that's not nothing. That's a very specific result pointing at something real. But — and here's where I want to pump the brakes a little — DunedinPACE is also the newest, least-validated of these clocks. So the question you have to sit with is: did the microbiome find the most sensitive instrument, or did it find the noisiest one?
Finn Brooks: Wait, so it might be a feature of DunedinPACE's design, not actual biology?
Juniper Vale: That's exactly the tension. We don't know yet. The 15% is real — it's statistically independent of chronological age, which matters — but 85% of your aging pace is still unaccounted for. Genes, environment, things we haven't mapped. So this is a piece.
Finn Brooks: Not the piece. Okay I was — yeah, I was maybe treating it like the piece.
Juniper Vale: And the study is 123 people, which for a proof-of-concept is fine, but it does mean we're not ready to hand anyone a microbiome prescription and call it longevity medicine. The signal is interesting. The causality questions are very much open.
Finn Brooks: But the causality thing — that's the one I can't shake. Like, the piece that keeps breaking my brain is: is the microbiome actually steering the ship, or is it just standing on deck watching everything happen?
Juniper Vale: That is the question. And there actually is evidence pointing toward causal — not just 'they travel together.'
Finn Brooks: The Mendelian randomization study. That's what I keep wanting to throw at this.
Juniper Vale: Walk through it.
Finn Brooks: Actually, wait, let me back into this — Mendelian randomization uses genetic variants, things you're born with, as like natural experiments. You didn't choose your microbiome-influencing genes any more than you chose your eye color. So if people with certain genetic variants that shape gut microbiota also age differently — healthspan, frailty — that's harder to explain away as confounding. And a 2024 paper in Journal of Translational Medicine did exactly that. Over a million individuals. MiBioGen consortium data. And it found causal links between gut microbiota and longevity outcomes.
Juniper Vale: Over a million. Yeah, that's not a small signal. But — and I'd push on this — that study is linking gut microbiota broadly to longevity. It is not telling you which bacteria, which intervention, which dose.
Finn Brooks: Right — so the causal arrow might be real but the map is still basically blank. Okay. What about the mouse study? July 11th, scFOS+ supplementation, 56 days, 18-month-old mice — that's like, old mice — they actually reversed some of the dysbiosis. Reduced the pathogenic bacteria load.
Juniper Vale: It's suggestive of reversibility, which matters. But think of it like this — imagine a 58-year-old woman in Honolulu gets her DunedinPACE result back and her clinician mentions the UH Mānoa cohort, mentions it included Native Hawaiian and Pacific Islander participants, people who look like her, and suddenly this research feels like it's actually about her aging, not some European ancestry dataset from the other side of the world. That moment is real. But what her clinician cannot yet tell her is whether changing her microbiome moves her DunedinPACE number. That's the gap.
Finn Brooks: And dysbiosis — the actual mechanism everyone's pointing at — reduced microbial diversity, loss of the short-chain fatty acid producers, more inflammation-associated bacteria — that pattern is consistent with causation but it's also completely consistent with the microbiome just being a passenger on a bus that's already crashing.
Juniper Vale: Exactly. And honestly — there's a whole other layer to why this is harder to resolve than it looks, something about the measurement infrastructure itself, that we should get into in a minute, because it makes the sample size problem even messier than it already sounds.
Finn Brooks: Wait — that measurement layer thing, that's the part that's been nagging at me. Like we can't even agree on how to read the data we already have?
Juniper Vale: Tanya T. Karagiannis and colleagues published in PLoS Computational Biology this year, looking at the Integrative Longevity Omics Study cohort — and what they found is that MetaPhlAn4 and Kraken2, the two dominant tools for classifying gut bacteria from stool samples, produce classifier-specific inferences. Same sample, different software, different answer about what's aging you.
Finn Brooks: Hold on. The same stool sample?
Juniper Vale: The same stool sample. And findings that one classifier surfaces, the other misses entirely. So if the UH Mānoa team had used Kraken2 instead of MetaPhlAn4 — or vice versa — we might be having a completely different conversation right now about whether the microbiome signal exists at all.
Finn Brooks: Okay that is — dude, that's not a footnote, that's a foundational crack. Like you can't build a clinical protocol on a measurement that shifts depending on which software your lab licensed.
Juniper Vale: Right — and stack that on n=123, which the authors themselves describe as proof-of-concept, and the honest verdict is: the signal is real, the 15% is not noise, but we have no longitudinal trials, no validated protocol, zero controlled human evidence that modulating your microbiome actually moves your DunedinPACE trajectory. The path from here to a healthspan intervention runs through all of that. None of it exists yet.
Finn Brooks: I mean — okay, I love that framing, BUT — I want to be precise about what we're actually saying. We're not saying the microbiome doesn't matter. We're saying we can't yet measure it reliably enough to act on it.
Juniper Vale: That's the calibrated take, yeah. The microbiome is probably a real, dynamic contributor to how fast you're aging — not just a passenger. But the measurement infrastructure isn't there, the sample is tiny, and causation from this study alone doesn't hold. Worth watching closely. Not worth rewriting your doctor's orders over.
Finn Brooks: So I keep thinking about — you know, I sent you that study at eleven at night because I thought it was going to be this clean 'the microbiome is the thing' story. And it's not that. It's more like... the field is naming the hypothesis before it can actually test it. Bautista and López-Cortés literally formalized a framework in their May 2026 Frontiers in Aging review — the 'microbiome-gerogene axis' — and called it a framework for precision geromedicine. They are naming a thing that doesn't have a clinical proof yet. That's either exciting or a little alarming depending on your mood.
Juniper Vale: Both, honestly. The question has shifted — and this is actually where I'll land — it's not whether we can measure a link between the microbiome and aging pace. The UH Mānoa study answered that, 15% of DunedinPACE variance, right there. The question now is whether we can reliably steer it in humans, long enough, to see any downstream effect on healthspan. That's a much harder problem. And nobody has done it.
Finn Brooks: So the take was right. Just maybe ten years early.
Juniper Vale: Eat your fiber. Know what you don't know.
