How RNAWiki checks
medicine evidence.

A medicine page starts with the intended use, the clearest research result and the biggest unanswered question. A general research summary can contain longer technical wording. RNAWiki keeps that wording available, but does not make readers decode it before seeing the main point.

The short answer must make sense without a glossary. If an unfamiliar study word is not essential, it moves out of the first screen. If it is essential, its familiar meaning appears in the same sentence. Readers do not need to hover or tap to understand the main point.

A separate “More about this medicine” section keeps useful general background, including safety information and how the medicine was given. Those details stay expandable and are not mixed into the short evidence conclusion or presented as personal dosing advice.

“See how we know” keeps the exact study names, measurements, comparison groups, source versions and professional wording for readers who want to inspect them. Simpler wording changes the reading order, not the underlying evidence.

On programme pages using the new presentation format, the mechanism map contains three to five ordered stages. Each stage has ordinary-language text, optional technical detail, a clear evidence label and links to the exact saved source versions behind its statements.

The evidence label says whether a stage was measured in people, measured only in laboratory or non-human work, predicted, or is not yet known. “Measured in people” requires a direct measurement or a regulator’s finding from a human study. A sponsor—the company or organisation running the study—may report a result, but that report alone is not labelled as a direct measurement. “Measured outside people” requires a direct measurement from a linked laboratory or non-human study. “Predicted” is a reviewer’s expectation, not a measurement; the software checks the linked statement and source, but it cannot verify an unstored model or reasoning.

A timeline appears only when reviewers have included real events that could change how the programme is understood. Every such event must point to the saved source version that supports it and say whether its date occurred, is planned, or was reported without clear timing. If there are no reviewed source events, RNAWiki hides the timeline instead of filling it with decorative milestones. Publication dates still appear in version history.

The map, timeline, sources and conclusion are locked into the same checked version before people review it. RNA Intelligence checks those links and labels; it does not write the map or decide the medical meaning.

This format is being added through reviewed versions for specific uses. Pages with only general or source information may not have a mechanism map or timeline, and RNAWiki does not invent either one to make coverage look complete.

RNAWiki is separating medicine research by intended use and group, one verified source at a time. It does not guess an intended use, dose, trial result, reviewer or conclusion to make a page look complete.

Current coverage: 8,987 medicine records, 0 development programmes recorded, and 0 with a published reviewed conclusion.

A medicine page without a reviewed answer for one specific question may therefore show a clearly labelled general research summary. Research found in an official registry may show its study and source but say that reviewers have not published a conclusion. Only evidence and a conclusion reviewed together are shown as a reviewed answer.

When a medicine page does not yet have research separated by its intended use, a reader can suggest a correction to one name at a time: either the medicine name or its trade or brand name. The contributor must give the exact replacement, the source page’s title and web address, and a plain explanation of why the change is needed.

Every identity correction waits for one independent human reviewer—even when the contributor is a trusted editor or administrator. People cannot review their own correction, and an accepted identity correction does not raise a contributor’s scientific-review standing. The reviewer checks the contributor’s source; RNAWiki does not fetch that page or claim that RNA Intelligence proved the name is true.

RNAWiki records the source, explanation, before-and-after values, contributor and database-set times together. These details cannot be changed after submission. At approval, the database checks that the old name still matches what the reviewer saw and applies that one change in the same step as the review decision.

Evidence, safety, effectiveness, trials, mechanisms and conclusions cannot use this small correction path. Those changes must be tied to one intended use, group of people, dose and set of studies, then checked by RNA Intelligence and reviewed in context.

Earlier pending edits that did not include these safeguards remain visible in the public history but are archived and cannot be reviewed or published. A contributor can submit a new sourced name correction or use the full evidence-review path instead.

RNAWiki has one account type. Every signed-in account can post a community note and submit the sourced correction or challenge available on a medicine page. Editorial standing and scientific-review qualifications control who may take part in review or complete publication later; they are permissions on the same account, not different kinds of login.

The server takes the author’s account identity and name from the signed-in session. A note or edit request cannot supply a different author. Community notes save the account attribution with the text. Edit proposals save the account identity, become unchangeable when submitted and show the contributor in the public review queue or history.

A community note appears as reader commentary, separate from the evidence record. RNA Intelligence does not check or fact-check it, and posting it cannot change the reviewed answer. An edit proposal follows the source, software-check and human-review paths described above; signing in never gives anyone direct publication access.

“Accepted contributions” counts programme proposals that reached the final accepted for implementation review state. It does not count requests for changes as accepted, does not mean every item has been published, and never raises an account’s trust level automatically. A public medicine record changes only after the separate publication review described above.

Reader feedback can include an optional contact address. Only a steward or administrator can see that private queue and close a report with a recorded reason. To limit repeated spam, RNAWiki stores a day-specific coded identifier for anonymous submissions. The original internet address is never stored. The coded identifier is not shown in the queue or any public response.

Every sentence says where it came from. A medicine page marks each statement as a reviewed conclusion, an answer a reviewer approved against this exact record, a sentence someone wrote into the record with the study named beside it, a quotation from a stored source, a count of rows RNAWiki holds, or a fixed RNAWiki sentence. An absence is shown with its reason rather than hidden.

Not measured, unknown, mixed, contradicted and confirmed are different states. Missing information is never filled in with generated medical content, and it is never turned into a negative result.

Anyone signed in may suggest better wording. The small “Review or improve” control at the top of a medicine page opens the review work for that page. Pick the sentence, write what you think it should say, say why, and attach the source if the change alters what the evidence shows.

Three members have to agree. Three different eligible members must approve the same exact wording before it replaces what is on the page. The person who proposed it cannot be one of them, editing a proposal starts the count again from zero, and one request for changes or one rejection ends it — on a medical page the safe answer to a disagreement is not to publish.

Some changes need a qualified reviewer. A change to what the evidence shows needs at least one of the three to hold a recorded qualification in a relevant field. A change that touches safety, interactions, pregnancy or how a medicine is used needs at least two. A reviewer who declares a conflict of interest can still say what they think; their decision does not count toward the three.

Agreeing is not enough. Deterministic checks run when a wording is submitted and again when the third approval lands. They refuse a change that turns an association into a cause, an animal result into a human one, a measurement into something a person would feel, or “no evidence found” into “safe”; that drops a limit, names an amount to take, tells a reader to start or stop treatment, or softens a serious warning. A failed check is shown to the reviewers by name and blocks publication.

Approval is about words, not about evidence. Three members agreeing changes how a sentence reads. It does not change what kind of evidence sits behind it: an animal result stays an animal result, a draft stays a draft, and a page does not become a reviewed page because a sentence on it was reworded.

Nothing is deleted. The wording that was replaced, the reason it changed, who approved it and whether a qualified reviewer took part all stay in the page’s own history. If a stored record changes under an approved wording, the approvals go stale and the page returns to what the record itself says rather than carrying a sentence nobody has checked against it. A steward can roll a published wording back, and the rollback is another entry in the history.

A correction to a stored record — a medicine’s name, a source, a trial, a conclusion — is a different path with its own review, described above. Rewording a sentence never edits the record it was built from.