Expert tier
Attributable, cited, ranked
Clinician protocols and named practitioner sources, graded high, medium, or low. Carries a citation back to the originating document. This is the only tier permitted to drive consensus and conflict detection.
Not open
to the public
The experiment
already ran.
We read the results.
Somewhere between a clinical trial and a rumour sits the largest body of real-world compound data that exists: what people actually took, at what dose, and what happened next. PepAxiom reads 1,658,809 of those accounts across a broad spread of public communities. 47,150 of them state a dose outright. Those are the ones a distribution can be built from, and the distance between those two numbers is the whole discipline.
Fig. 1
Reported dose distribution, one compound, one indication
In the
ledger
Eleven numbers, none of them rounded up
Counts as of the last consolidation pass. Every one is read straight out of the engine; none is an estimate, and none is a ceiling we are working towards.
- Anecdotal records read
- 1,658,809
- Records stating an explicit dose
- 47,150
- Compound-level mentions extracted
- 602,865
- Records resolved to a compound
- 243,715
- Compounds catalogued
- 1,231
- Under continuous tracking
- 76
- Compounds with a dose distribution
- 67
- Protocols from vetted sources
- 469
- Synthesised compound dossiers
- 204
- Compound-to-biomarker links
- 545
- Co-use pairs mapped
- 794
Method
How a rumour becomes a record
Three passes, in order. A claim that does not survive all three does not enter the ledger.
-
i Collect wide
We read across a broad spread of public communities where people compare notes in the open. No single platform, no single vocabulary, no single demographic. Each corpus is kept provenance-distinct from collection through to display, so a signal that only exists in one place can never masquerade as consensus.
-
ii Resolve to entities
People do not write in schemas. Names arrive as slang, misspellings, brand names, blend nicknames, and abbreviations that collide with unrelated compounds. Every one is resolved to a canonical entity, and doses are parsed out of ordinary prose, including the reconstitution arithmetic people do in their heads.
-
iii Weigh, then separate
Every record carries a credibility tier that follows it everywhere. Volume is counted but never promoted into authority. Where the crowd and the clinical literature disagree, the disagreement is the output. We surface the conflict rather than averaging it into a number that describes nobody.
The load-bearing rule
Volume is not evidence.
Most tools that read community data pour everything into one average and hand you a number. That number is worse than nothing, because it looks like an answer. In PepAxiom, expert-sourced evidence and community-aggregate consensus are stored separately, scored separately, and rendered on surfaces that never touch. You always know which one you are reading, because they do not look alike.
Records read
1,658,809
Resolved to a named compound
243,715
Also stating an explicit dose
47,150
Which is enough to build a dose distribution on
67 of 1,231 compounds
Ninety-seven per cent of what we read never becomes a dose figure, and we would rather show you that than the headline. The corpus size is not the sample size: every distribution carries its own n, and where that n is thin we say so on the compound rather than letting the number at the top of this page speak for it.
Two tiers
Community aggregate
Counted, never promoted
What large numbers of people independently report doing. Genuinely useful, and genuinely not the same kind of thing. Rendered in its own visual register so it can never be mistaken for clinical guidance, no matter how large the sample gets.
Behind
the gate
Most of this is not public
What is on this page is the outline. The engine itself is internal, and stays that way for now.
- Compound dossiers Restricted
- Reported dose distributions Restricted
- Conflict detection Restricted
- Biomarker panels Restricted
- Stack co-use graph Restricted
- Protocol library Restricted
- Longitudinal signal monitor Restricted
Restricted is not the same as unpriced. The tier at which each of these opens, and what it costs, is on the price list.
Request
access
Tell us what you would do with it
Access is granted case by case, mostly to clinicians, researchers, and people building something serious on top of it. A real sentence about your work gets a real reply.
- Reply
- From a person, not a sequence
- Access
- Read-only, scoped to what you need it for
- Your details
- Not stored here, not passed on
Currently not accepting
new clients
The form will reopen when capacity does.