The atlas started from a simple observation: the peptide information landscape is dominated by commercial catalogues, forum posts, and social-media claims, while the primary evidence — FDA labels, EMA public assessment reports, ClinicalTrials.gov registries, PubMed-indexed studies — is scattered across dozens of databases with different access patterns and update cycles. A researcher trying to answer "Is this peptide approved for anything?" or "What human evidence actually exists for this claim?" could spend hours crossing sources without finding a clear answer.
This post explains how the atlas was built, what it covers, and the rules that keep it source-traceable. The underlying methodology is documented across Evidence grading, Scope and selection, Dose language, and Identity and structure.
Scope: 100 entries, four classes, boundary cases explicit
The Scope and selection policy defines four non-equivalent inclusion classes: regulator-approved peptide active substances, clinical-stage investigational peptides, endogenous or laboratory peptides with substantial scientific visibility, and products marketed in the "research peptide" ecosystem. Inclusion reflects scientific, clinical, regulatory, or public visibility — not endorsement.
The Coverage report records the result: 100 catalog entries, 100 monographs present, 118,893 total monograph words. The entries span eight categories — antiinfective/coagulation (13), cosmetic/other (11), GI/endocrine/cardiovascular (13), growth axis (13), melanocortin/reproductive (12), metabolic/incretin (13), neuropsychiatric (12), repair/mitochondrial (13) — and five entry types: approved drug (58), approved diagnostic (1), investigational (13), research market (15), endogenous (6), and boundary case (7).
Boundary cases matter. As the Identity and structure page explains, some entries sold under a single name are not a single well-defined substance: TB-500 is ambiguous between full-length thymosin beta-4 and a fragment, GHK-Cu's PubChem record represents a bis-complex not the generic 1:1 species, and thymalin and cerebrolysin are undefined peptide mixtures from tissue extracts with no single CID or sequence. The atlas classifies these explicitly rather than assuming that everything sold as a "peptide" is one.
Evidence methodology: grades attach to claims, not molecules
Every monograph follows the template defined in AGENTS.md: exact identity and aliases, class and sequence, a status table separating jurisdictions, mechanism with evidence qualifier, indication-by-indication evidence table, approved-label regimens plus studied exposures, safety signals, regulatory and WADA status, and evidence gaps.
The Evidence grading page defines the A–E and X scale, but the key methodological choice is that the grade attaches to a specific claim, not to the molecule as a whole. The tirzepatide monograph demonstrates this: grade A for glycemic control in T2D, grade B for OSA with obesity, because the claims have different regulatory and evidence status. The BPC-157 monograph shows the same principle at the other end of the scale: grade C for its interstitial cystitis pilot (small, single-arm, uncontrolled) and grade D for musculoskeletal healing (preclinical evidence only).
Source rules: primary over secondary, dates over promises
The repository's README establishes the citation standard: every material claim should cite a stable primary or authoritative source — regulator labels and safety communications, trial registries, peer-reviewed human studies, PubMed records, recognized chemical databases, pharmacopoeial standards, and official statutes or agency guidance. Narrative reviews may orient a page but may not be the sole support for efficacy, safety, dose, or legal-status claims.
The AGENTS.md source rules add: third-party blogs, clinic pages, social posts, and product listings may establish that a claim or blend is marketed; they cannot establish efficacy, safety, purity, dose, or legal status. Never invent a DOI, PMID, NCT number, label statement, sequence, approval, or jurisdictional rule. A URL returning successfully does not prove that the source supports the claim; reviewers must inspect decisive sources.
The Dose language policy operationalises the safety boundary: "recommended dose" is reserved for regulator-approved product information naming jurisdiction, indication, population, route, and label date. All other exposures are reported as "studied regimens." If no approved human use exists, the monograph states: "No established or recommended human dose."
What the coverage report shows and does not show
The generated coverage report measures presence — 100 of 100 monographs present, zero sparse pages or missing pages — but it does not measure scientific validity. A monograph can be complete, well-sourced, and still contain a claim that future evidence overturns. That is why every time-sensitive record carries an as_of or last_verified date and should be rechecked before publication or decision-making, as the Scope and selection refresh policy requires.
The retatrutide monograph is a useful illustration: its entry_type is "investigational," it has no marketing authorization, yet its evidence table carries grade B human evidence from controlled Phase 2 and 3 trials. The distinction between development-stage evidence and regulatory approval is preserved in separate tables, and neither is allowed to collapse into the other.
The meta point
The atlas is not a static reference. It is a living collection designed to be checked, challenged, and updated as new trials post results, labels change, and regulatory decisions shift. The methodology pages exist to make the rules of the review transparent — so that every reader can see not just what the atlas concluded, but how it got there.
The full methodology documentation starts at Evidence grading.