Dataset
Derived evidence tier, per compound
One row per compound: the derived evidence tier, the basis sentence that produced it, the record counts by study model behind that basis, and whether the reference publication currently publishes the compound's page as indexable.
77 rows13 columnsGenerated 13 October 2026
The file
77 rows, 13 columns, 14 KB. Comma separated, UTF-8, one header row, quoted where a value contains a comma or a quote. Download evidence-tier-mix.csv.
How it was derived
Read straight out of data/evidence/<slug>.json, which is generated by scripts/build-evidence.mjs from counts of record types. The tier is assigned by an ordered ladder of tests and the first match wins; the basis column states the counts that matched. Nothing in this file involves reading a study.
Regenerate it with node scripts/sentinel-analysis.mjs --write. The same run produces every figure in 30 of 77 carry no human-trial record, so the page and the file cannot drift apart.
Columns
| Column | What it holds |
|---|---|
| slug | The compound's identifier in this corpus. |
| name | Display name as the corpus holds it. |
| tier | The derived label. One of human_trial_evidence, limited_human_evidence, animal_and_in_vitro_only, in_vitro_only, preparation_or_mixture, model_not_established, sparse. |
| classification | Always INFERRED. The tier is derived metadata, never a measured value. |
| basis | The sentence stating which counts produced the tier. This column, not the label, is the checkable part. |
| indexed_rows | Records the corpus carries for this compound. |
| human_trial_rows | Of those, how many carry the subject heading Humans together with a trial publication type. |
| rodent_rows | Of those, how many carry a rodent subject heading. |
| in_vitro_rows | Of those, how many carry an in vitro subject heading. |
| other_rows | Of those, how many carry no heading that settles a study model. In this corpus this is the largest category. |
| review_tier | routine or high_impact. high_impact means the name carries a drug application record, or the set contains a trial record, or the compound sits in a weight-management or metabolic class. |
| qualified_review | yes if a named qualified reviewer is recorded against the profile. |
| withheld_from_indexing | yes if the reference publication builds the page but holds it back from indexing, which is every high_impact profile without a recorded reviewer. |
What it cannot be used for
The tier describes what this corpus holds about a compound. It is not an assessment of clinical efficacy, of safety, of study quality or of scientific consensus, and it must not be read or re-published as one.
The limits of the analysis built on this file are stated on the investigation itself, and the rules this publication applies to all of its counting are on the methods page.