Download
- Full dataset as JSON — every compound record, CORS-enabled for direct use.
- Compound classes as JSON — the taxonomy: every pharmacological family with its members.
- Plain-text digest — the whole reference as text, the cheapest representation for a language model to read.
- llms.txt — the index of what is here.
Version 2026.10.07 (schema 1.2.0) — the version is the newest review date in the data, so it moves with the content · last reviewed 7 October 2026.
What is in a record
Every field is documented so that a value can be interpreted without reading the site. Grades count controlled human evidence only — animal and cell studies never lift a grade above D. Effect size is recorded separately from grade, so a large uncertain effect and a small certain one are never conflated.
idstringURL slug, stable identifier for the compound.namestringDisplay name of the compound.aliasesstring[]Synonyms, brand names and research codes.urlstringCanonical LYO page for the compound.classobjectThe pharmacological family the compound belongs to (id, name, url). Every compound is in exactly one; the full taxonomy is in the classes file.summarystringAnswer-first paragraph describing what the compound is and what the evidence shows.overall_gradeA|B|C|DThe best-supported outcome's grade. A established in humans, B probable, C weak, D no human evidence.human_trialsstringShort description of the controlled human trial base.regulatory_statusstringShort approval-status label.moleculeobjectMolecular identity from PubChem: formula, molecular_weight (g/mol), cas, pubchem_cid; note explains records that have none (proteins, mixtures, combinations).pharmacokineticsobjectRoute, half_life and, for approved products only, the licensed_regimen from the label. basis says whether the figures come from a label, a human study, animal data or nothing.outcomes[].outcomestringOne claimed or observed effect.outcomes[].gradeA|B|C|DEvidence grade for that outcome. Counts controlled human evidence only.outcomes[].effect_sizenone|small|moderate|largeMagnitude of the effect where one is demonstrated; independent of the grade.outcomes[].adversebooleanTrue when the outcome is a harm rather than a benefit.outcomes[].human_studiesstringThe human study base for that specific outcome.outcomes[].evidencestringPlain-language summary of what the studies found.outcomes[].referencesobject[]Citation, venue and URL for each primary source behind the outcome.outcomes[].verbatimobjectA sentence copied verbatim from the cited abstract, with its citation and URL. Present only where a quote was verified against the source.regions[]object[]Regulatory status per jurisdiction (EU, US, UK, Australia, Canada, sport/WADA), with a note.faq[]object[]Question-and-answer pairs published on the compound page.references[]object[]All primary sources for the compound, numbered as shown on the page.last_revieweddateDate the compound record was last reviewed.literature_searched_todateDate the literature search behind the record was last run.changelog[]object[]Dated revision history for the record.Licence and citation
The dataset is published under Creative Commons Attribution 4.0. You may copy, adapt and redistribute it, including commercially, provided you credit LYO Research and link back.
LYO Research (2026). LYO Research Peptide Evidence Database, version 2026.10.07. https://lyoresearch.com/data
How to read the grades responsibly
A grade describes the state of the human evidence, not safety, not legality, and not whether something is appropriate for anyone. A D does not mean a compound does not work — it means no controlled human trial has shown that it does. An A on an adverse outcome means a side effect is well established, which is not a recommendation either way. If you reuse the grades, carry that distinction with them.
The method behind every grade is public: see how LYO grades evidence and how pages are written and checked.
Found an error?
Corrections are welcome from anyone and are made publicly with a date. A record, a field and the source that contradicts it is the fastest possible fix — send it here.