Method 01
Find the source
We start with the original research whenever we can—not another website repeating it.
Our method
We don't ask you to
take our word for it.
A clear look at how OnimaChain finds sources, checks records, keeps different kinds of evidence separate, and tells you when something is missing.
Method 02
Check the record
Study titles, publication details, and identifiers are verified before we show them as confirmed research.
Method 03
Keep stories separate
What researchers measured and what people say online are both worth exploring—but they aren't the same kind of evidence.
Method 04
Show the gaps
If we haven't checked something yet, we say that. If research stops at animals, we show where it stops.
Go deeper
See exactly how this works.
The plain-language view comes first. The underlying process stays available for researchers and careful readers.
How sources enter the library+
Scientific records enter through PubMed identifiers in a versioned data file. Every identifier is checked against NCBI before the site builds. Titles, dates, journals, and author lists come from that record—not promotional copy. Last corpus check: 2026-09-22T22:40:47.232Z.
How we connect claims to research+
Every tracked claim has an origin record or a clear “connection not yet verified” state. We preserve the original scope, including the species, model, endpoint, and source wording. Later interpretations stay visibly separate.
How we handle real-world reports+
No community-source connector is enabled yet, so we don't display invented totals. When collection begins, reports will retain platform, date, research goal, co-interventions, and duplication relationships.
How we look for independent research+
We currently show distinct last-author names as a limited proxy. Institutions, funding, and citation networks remain unchecked until the necessary source connectors are available.
Where people review the record+
Study-to-claim relationships are editorial decisions with a source basis. The data files are currently the review surface, and changes are versioned in git. A dedicated editorial console is planned.
How corrections work+
The public corrections ledger records the date, affected record, before-and-after text, reason, and source. Previous interpretations are superseded, not quietly erased.
Known limits+
The current corpus contains 9 verified records across two compounds. No trial, regulatory, or community connector is enabled. Three-dimensional structures are procedural, sequence-derived visualizations—not measured structures unless explicitly identified as deposited.
Technical implementation
The record underneath the interface.
Study { pmid, compounds[], species, studyType, tags[], sourceQuote?, status, NCBI metadata }
Claim { id, title, compound, originStudy|null, originalScope, mutation[], support[], translation, contradictions[], changeHistory[] }
Relations study SUPPORTS | PARTIALLY_SUPPORTS | CONTRADICTS | DOES_NOT_TEST claim
Structure provenance ∈ { deposited PDB | sequence-derived visualization | conceptual visualization }