Peplexicon evidence method · versioned and inspectable
Peplexicon is an evidence system, not an oracle. It preserves source history, attaches context to claims, publishes explicit coverage boundaries, and changes its interpretation when the evidence changes.
Research loop
The pipeline separates discovery from publication. Search results and research leads can tell the system what to inspect, but they cannot directly become public evidence.
Scheduled adapters check registries, publications, labels, and research leads. A failed fetch is recorded as a failure—not as ‘nothing changed.’
Changed upstream records become immutable, content-hashed snapshots. Earlier versions remain addressable.
The compiler proposes scoped claims with exact source passages, locators, study context, and evidence direction.
Deterministic assurance rules check source authority, independence, applicability, ambiguity, conflicts, corrections, and retractions.
Eligible claims are assembled into a deterministic publication. A semantic change creates a new immutable version.
The public profile advances to the verified version. Withheld claims remain outside the published evidence surface.
Publication states
Publication state describes process, not whether a result is favorable. Source count, evidence strength, profile completeness, and reviewer state are separate signals.
The peptide exists in the directory, but its page may not yet have a published claim ledger.
Published claims passed the current deterministic policy and resolve to validated evidence snapshots. No human review is asserted.
A named qualified reviewer approved an exact evidence input hash. This state is never inferred from automation.
Evidence controls
Multiple articles can report the same trial, and a large study can still be poorly applicable to a specific claim. Peplexicon therefore preserves both citation-level provenance and study-level independence.
Population, intervention, comparator, endpoint, formulation, dose, and time horizon prevent a narrow result from becoming a universal claim.
Edges can support, qualify, contextualize, or contradict. The current profile does not erase an inconvenient record.
Registry entries, primary papers, and secondary analyses can resolve to one underlying study instead of inflating consensus.
New label versions, corrections, retractions, and changed trial records can invalidate dependent assessments and trigger a new profile version.
Coverage
A published profile is a validated subset of the compiled research surface. It may withhold claims because evidence is weak, ambiguous, high-risk, contradictory, stale, or not yet supported by an authoritative source.
Inspect the system
Use the public operations page to inspect freshness, scheduled-run proof, publication delivery, and current failures. Open a profile claim to inspect its evidence edges and source passages.