Public datasets
Inspect what this corpus records—and what it does not.
These six read-only views publish deterministic measurements of RNAWiki’s recorded corpus. Each one states its scope, method, source boundary, and limitations before the rows.
Six allowlisted projections
Choose a dataset
- 8,404 rows
Enzyme and transporter findings
Inspect every admitted source sentence for an exact medicine, enzyme or transporter, and recorded role, with assertions, denials, and unresolved polarity counted separately.
This is not a drug interaction checker. A denial applies only to the named medicine, role, counterparty, conditions, and source sentence; it is not proof of no interaction and is not dosing or treatment advice.
Generated
Inspect1.3.0 - 1,681 rows
Source agreement and disagreement
Show whether sources print one normalized reading, have incompatible units, lack structured comparison context, or contain a context-matched numeric difference.
agree can mean only that sources print the same normalized reading; it is not clinical-context equivalence, independent replication, or proof. differ requires matching structured context and disjoint comparable values. not_comparable and insufficient_context are not disagreement.
Generated
Inspectsource-consensus/2 - 167,535 rows
Silence and non-establishment
Show, for each medicine and tracked question, whether the recorded source sections contain an answer, explicitly say the question is not established, or stay silent.
SILENT means only that this corpus did not record an answer from the sections it read. It is not evidence of safety, danger, or real-world absence. NOT_ESTABLISHED is an explicit source state and must not be collapsed into silence.
Generated
Inspect1.2.0 - 9,855 rows
Corpus coverage
Show which recorded-background modules, source-bound ordinary answers, specialist measures, conflicts, and explicit question gaps are observable for each medicine in the checked-in public snapshot.
A larger module count does not mean a medicine is better studied, safer, more effective, or higher quality. Different source routes can supply different kinds of fields and their counts are not directly comparable.
Generated
Inspect1.4.0 - 9,859 rows
Record identity resolution
Look up what one stored record was decided to be: an entity that keeps its own address, a record that resolves to the address of the same entity, or an identifier that never named a medicine.
This is about records, not about substances. It does not say two substances are the same thing, that a record is correct, or that a withdrawn address means the substance does not exist. A registry identifier held by more than one record is published as a warning code and is never treated as evidence to merge.
Generated
Inspectinventory-resolution/1 - 9,852 rows
Dossier completion states
Read, for one canonical record, every section that applies to it and the state that section reached over the sources RNAWiki holds.
Every state describes the sources that were read, never the medicine. "No qualifying evidence after search" is a statement about this corpus and its dated searches, not a safety finding and not an absence of risk. A complete record is one whose applicable sections all carry an explicit state, not one that holds a positive result in every section.
Generated
Inspectdossier-completion/1
Access and reuse
Bounded API
JSON and CSV responses expose only named public fields, at most 200 rows per request.
Server-side projection
Large source artifacts are parsed on the server. The browser receives only the requested page.
Visible boundaries
No private records, review-work payloads, generic file paths, or automated medical conclusions are served.
Licence and source rights
RNAWiki’s selection, schema, structure, and derived projections are licensed under CC BY 4.0. Credit RNAWiki and the generated snapshot date.
Quoted source passages and third-party records retain their original rights. A dataset licence does not transfer ownership of an FDA label, paper, registry record, or other cited source.