Loading
Two to four articles, one column each, on the three questions that separate an approved therapy from a promising mouse result: what has been shown in people, who can obtain it, and what happens if it turns out to be wrong. Everything shown is drawn from the articles themselves.
3 of 4 · minimum 2
Arrow keys move through the list, Enter adds, Backspace on an empty field removes the last chip.
In all 3 articles
Scaled to the full 1–9 TRL scale rather than to the selection, because TRL 4 next to TRL 5 is a small gap and normalising would make it look large.
Five ordered positions from reversible on the left to irreversible on the right. Further right is harder to undo, which is not the same as worse. Describes the intervention as the article scopes it, and assumes it was delivered as intended. Reversing an effect is not the same as undoing its consequences.
| Field | CRISPR–Cas9 | Base editing | Prime editing |
|---|---|---|---|
| Portal | Genetics | Genetics | Genetics |
| Type | Technology | Technology | Technology |
| Evidence | Established | Emerging | Experimental |
| In humans | One approved medicine, Casgevy, edits a patient's own blood stem cells outside the body; in vivo editing has durably lowered the disease-causing protein in transthyretin amyloidosis, and heritable editing is unapproved in every country with a relevant law. | All human use is early-phase or one-off: base-edited donor T cells were given to a teenager with relapsed leukaemia in 2022, in vivo liver editing has lowered cholesterol in early trials, and a bespoke editor was dosed into one infant. | Human use is one ex vivo trial in chronic granulomatous disease, which reported restored immune-cell function in a treated patient in 2025; in vivo correction has been shown only in mice. |
| Horizon | present | late 2020s | 2030s |
| Readiness | TRL 9 | TRL 7 | TRL 6 |
| Reversibility | Irreversible No known route back. Includes anything heritable. | Irreversible No known route back. Includes anything heritable. | Irreversible No known route back. Includes anything heritable. |
| Access | Reagents are ordinary catalogue items sold to any laboratory; the one approved medicine built on the platform treats two inherited blood disorders and is priced in the millions of dollars per patient. | No base-editing therapy is approved anywhere as of 2026; the only routes are early-phase clinical trials and one-off bespoke treatments authorised for a single patient. | Not obtainable as a therapy: no prime editing product is approved anywhere as of 2026 and human use is confined to early-phase trials; the reagents are sold for research. |
| Tags |
|
|
|
| Summary | A bacterial defence system repurposed as a programmable DNA-cutting tool, in which a short guide RNA directs the Cas9 enzyme to a chosen sequence in almost any genome. | A genome editing method that chemically converts one DNA base into another at a targeted site without cutting both strands of the double helix. | A genome editing method in which a nicking Cas9 fused to a reverse transcriptase writes a new DNA sequence specified by an extended guide RNA. |
| Article page | Read 1,763 words · 41 inbound links | Read 1,200 words · 19 inbound links | Read 1,259 words · 16 inbound links |
Every value in the table is read from an article's own frontmatter or computed from its text. Nothing is scored, weighted, or ranked by this page. The rows appear in the same order as the infobox on an article, so a reader arriving from one does not have to learn a second layout.
Evidence status is editorial and assigned per article: established, emerging, experimental, speculative, contested, or historical. Horizon is a rough answer to when the topic plausibly matters at scale. Both are judgements, and both are stated on the article they come from.
In humans is one sentence on what has been demonstrated in people, as distinct from in animals or in cells. It is the wiki's central editorial rule made into a field, and it exists because prose honours that distinction unevenly. Read beside the evidence chip it qualifies: “experimental” and “no person has been treated” are different claims, and the second is the one that stops a mouse result being read as a human one.
Access is one sentence on who can obtain the thing today and on what terms — availability, legality, and cost. It is a statement about the world, never a recommendation, and this site gives no medical advice.
Reversibility is the axis readiness cannot express. A daily pill at TRL 9 and a heritable edit at TRL 3 differ on what happens if the science turns out to be wrong, not on how far each has been built. Five ordered values, from reversible to irreversible; anything heritable is irreversible.
These fields are absent by design on people, organizations, and completed events, and the taxonomy gates each one separately: human evidence is carried by technologies, interventions, and risks; access and reversibility by technologies and interventions only. Evidence status and horizon are withheld on the same three types, because an evidence grade judges a claim and a person is not a claim.
Where a field does not apply, the cell says so and gives the reason rather than sitting blank. A blank cell reads as “nobody has checked”, which is a different and worse answer. A cell reading “not stated” is the genuine gap: the field applies to that kind of article and has not been filled in. What an article actually says always wins over what its type would predict — a few articles answer one of these questions on a type that is not required to, and the answer is shown.
The same logic covers readiness. Concepts, people, and organizations show “not applicable” rather than a zero, because an absent TRL means the scale does not apply, not that the readiness is low.
The readiness bars are drawn against the full 1–9 scale. Normalising them to the largest value in the selection would make a one-level gap fill the whole track, which reads as a much bigger difference than it is.
The reversibility strip beside it is not a bar. A bar that fills further to the right says more, and more of this is not better — only harder to undo. The whole five-point scale stays visible and the article's position is the one segment shown at full strength, so an article with no position on the axis reads as absent rather than as zero. The two panels are the same pair of axes that the readiness page crosses across the whole corpus.
Word count and inbound links have not been dropped, but they are facts about the article rather than about its subject, so they sit under the link to the page instead of in rows of their own. Inbound links are counted from the wiki's link graph: the number of other articles that point here.
The selection is written into the page fragment as #c=slug,slug. Copy the address bar and the same comparison reopens. The fragment is read once on load; after that, changing the selection rewrites it without adding a history entry.
Tags are hand-assigned per article and reused deliberately across the corpus, so overlap is a real signal that two topics sit in the same neighbourhood. It is not a claim that the technologies are substitutes. Two articles can share every tag and still describe approaches that rule each other out.