Personal Science · Field Notes
A microbe earns its keep only when the target sits in the gut lumen — where it can act before or instead of the host — and there's a real path from “moved the biomarker” to “changed the outcome.” Most ideas die on one or the other.
| Target / capability | How it works (niche) | Pros | Cons & skeptic's flag | Verdict |
|---|---|---|---|---|
| Gout / high uric acid | Gut bacteria degrade urate → xanthine/SCFAs; patches the lost uricase gene | Uricase is a human pseudogene — a clean “lost-gene patch”; the gut already disposes ~⅓ of urate; large population; mouse causal data | No controlled human urate-lowering trials yet; won't replace xanthine-oxidase inhibitors; urate also handled renally | Strongpreclinical → early · the GULO twin |
| Kidney stones / oxalate | O. formigenes degrades dietary oxalate in the lumen, before absorption | Carriers ~70% lower recurrent-stone risk; colonization lowers urinary oxalate; clear luminal logic | Only ~half of trials significant; Phase 3 (Oxabact) missed its primary endpoint (p = 0.064); a hub species in a network; hard to colonize | ModeratePhase 3 · mixed |
| Cavities (Lantern / Lumina) | Engineered acid-free S. mutans permanently replaces the wild strain in the oral biofilm | Hits the root cause; one-time colonization; the oral niche is accessible | Potentially irreversible self-experiment; ecological & safety unknowns; off the standard trial path | Real · cautioncommercial · oral niche |
| PKU / phenylalanine (Synlogic) | Engineered E. coli Nissle expresses PAL, consumes dietary Phe in the GI; transient, non-colonizing | Proof-of-mechanism solid (~20–34% blood-Phe drop); strong biocontainment design | Phase 3 (Synpheny-3) discontinued Feb 2024 for likely futility; company wound down; needs continuous dosing | Mechanism, not clinicthe cautionary tale |
| Acetaldehyde / ALDH2 “flush” | Strain expressing aldehyde dehydrogenase mops up local acetaldehyde | Clean target for ~½ of East Asians; reduces a carcinogen | Ethanol is absorbed fast upstream; benefit likely limited to local gut/oral acetaldehyde | Niche · plausibleresearch-stage |
| TMAO / cardiovascular | Non-producing or TMA-consuming strain lowers choline/carnitine → TMA → TMAO | Diet-driven luminal substrate; on-trend | TMAO's causal role is contested — you'd be optimizing a disputed proxy | Speculative |
| Histamine intolerance | Histamine-degrading strain lowers dietary histamine in the gut | Luminal substrate; a defined sufferer group | Fuzzy diagnosis (DAO deficiency poorly defined); thin evidence | Niche |
| Heavy-metal burden | Lactobacilli bind lead / cadmium / mercury in the lumen, cutting absorption | Genuinely new capability; cheap | Mostly rodent / in-vitro; binding capacity vs. real exposure unclear | Early |
| Gluten / celiac | Prolyl-endopeptidase strain degrades the immunogenic gliadin peptides | Large demand | The caffeine problem — must act fast pre-absorption, and celiac reacts to trace amounts; enzyme versions underwhelm | Kinetically hard |
| Caffeine clearance | (proposed) gut microbe degrades caffeine | Enzymes & precedent exist (coffee-borer beetle; 71+ degrading strains) | ~99% absorbed upstream; ~95% cleared hepatically by CYP1A2 — the colon sees only crumbs | Fails · locationreframe: upstream enzyme pill |
| Vitamin C synthesis | (proposed) microbiome makes ascorbate | — | It's a host hepatic gene (GULO), multi-exon-deleted; microbes can't deliver it to the host; little benefit over diet | Fails · wrong systemthat's gene therapy |
| B12 status | Gut bacteria synthesize B12 | — | Made in the colon, downstream of the ileal absorption site — so it isn't absorbed | Fails · locationthe instructive counterexample |
These functions ride in the microbe, not your genome — so there's no “ethnic incompatibility.” The barriers are colonization and substrate, not ancestry.
| Module / origin | What it does | Repurpose play | Catch | Verdict |
|---|---|---|---|---|
| Equol from soy — Asian / vegetarian | Gut bacteria (Eggerthella, Slackia, Asaccharobacter) convert daidzein → equol, a potent ERβ modulator | An equol-producer probiotic or S-equol supplement for the ~70% of Westerners who can't make it | Equol's benefit is still partly correlational; needs a soy substrate | Strong wedgepayoff = a specific metabolite |
| HMO digestion — B. infantis | Specialist at human-milk oligosaccharides; depleted in many industrialized-country infants | Probiotic (e.g., EVC001) to restore HMO use in breastfed babies | Substrate-locked (needs breast milk); infant-only | Real product |
| Seaweed CAZymes — Japanese gut | Porphyranases/agarases (acquired by HGT from marine Zobellia into B. plebeius) digest nori (海苔) glycans | Hand the porphyran-utilization locus to non-seaweed populations to ferment kelp/nori | Useless without seaweed in the diet; engraftment; upside ≈ marginal extra fiber | Real · low-valuesubstrate-dependent |
| Lactose — colonic adaptation | The colon microbiome adapts to lactose with regular exposure (host LCT gene unchanged) | Gradual exposure / probiotic in lactase-nonpersistent adults | It's adaptation, not a new enzyme — modest effect | Modest |
Thinner evidence, same idea: dairy-adapted microbiomes (Mongolian, Maasai), the Hadza's fiber-degrading consortia, cassava cyanogen detox, tannin degraders, arsenic-methylating microbes in rice-eating populations.
A probiotic is worth pitching only when (1) the target sits in the gut lumen, where a microbe can act before or instead of the host, and (2) there's a credible route from a moved biomarker to a changed outcome. The bottom of Table I fails on location. The engineered living-medicine programs (Synlogic, Oxabact) cleared bar (1) and moved the biomarker — then missed the clinical endpoint, which is where this whole category keeps dying. Knowing why a concept fails is the sharper move than the concept itself.