要約
The guide finds its target by matching about twenty letters, but the match does not have to be perfect. Somewhere else in three billion letters there may be a sequence close enough to fool it. Editing there is an off-target edit — and because the change is permanent, it stays. This is measured, not assumed: it is one of the things regulators look at most carefully.
Off-target editing occurs where a guide tolerates mismatches, particularly PAM-distal ones. It is assessed by genome-wide empirical methods — GUIDE-seq, CIRCLE-seq, DISCOVER-seq — combined with targeted deep sequencing of nominated sites in the edited product. Mitigation includes high-fidelity Cas variants, truncated guides, transient ribonucleoprotein delivery and, most fundamentally, choosing a target site with few genomic near-matches.
How much does it matter?
It depends entirely on where the off-target edit lands. Most of the genome does nothing in any given cell type, and a change there is likely harmless. An edit in a tumour suppressor gene is a different matter. This is why the question is never 'is there off-target editing' — there almost always is some — but 'where, how much, and does it matter in this tissue'.
It is also why delivery method matters for safety and not only for logistics: ribonucleoprotein delivered directly degrades within a day or two, while DNA encoding the editor may keep producing it for weeks, and every extra hour is more opportunity to cut the wrong place.
Try it: find a target the way a scientist would
Editing is not "point at a gene". The enzyme can only cut where a short tag sits beside the target, so the first job is finding legal positions. Paste any DNA sequence — or use the example — and this will scan both strands for you.
This is a teaching model of the first step only. A real design run also searches the whole genome for near-matches that could be cut by mistake, scores predicted efficiency, and is then confirmed by sequencing in cells — none of which can be done from a short sequence alone. The example is an illustrative sequence, not a real genomic locus.
Sources
- National Human Genome Research Institute
Talking Glossary of Genomic Terms ↗
Check your understanding — CRISPR basics
No score is stored and nothing is sent anywhere — this is just for you.
1. Which part of CRISPR-Cas9 do you actually change to target a different gene?
Why: Cas9 stays the same; the ~20-letter guide is the programmable part. That is the whole reason the technology spread so fast.
2. What is a PAM?
Why: No PAM, no cut — which means some positions in a gene simply cannot be reached with a given enzyme. It is a real constraint on what is targetable.
3. After Cas9 cuts, what most often happens?
Why: Rough end-joining dominates. That is why switching a gene OFF is reliable while correcting one letter is hard — and why base and prime editing were developed.
4. What does 'off-target' mean?
Why: Similar sequences elsewhere in the genome can be mistaken for the target. It is found by sequencing, and it is one of the main safety questions regulators ask about.