The Cas9-gRNA complex slides along the DNA helix, checking each PAM (NGG) site it passes for a match to its 20-nt guide sequence. A perfect on-target match binds and cuts almost every time; a partial off-target match only binds if the guide's fidelity is loose enough to tolerate the mismatches. Every double-strand break is then resolved by one of two repair pathways racing inside the cell.
P(bind) = target ? ~0.95 : matchScore · (1 − fidelity)
P(HDR) = donorPresent ? ~0.45 : ~0.03
P(NHEJ) = 1 − P(HDR)
- Search speed — how fast the complex scans the strand; higher speed means more PAM sites are sampled per second.
- Guide RNA fidelity — higher fidelity rejects mismatched (off-target) sites; lower fidelity lets the complex bind loosely-matched decoy loci too, exactly like a poorly-designed sgRNA.
- Donor template — supplying a homologous repair template raises the odds of precise Homology-Directed Repair (green); without it, Non-Homologous End Joining (orange) dominates and leaves a small insertion/deletion.
- Editing efficiency — the share of the true target loci (green rings) that have been successfully cut and repaired so far, the same metric used to grade a real guide RNA design.
Real-world relevance: this bind → cut → repair race is exactly what determines whether a therapy like Casgevy edits only the intended gene or also disturbs look-alike sequences elsewhere in the genome — off-target risk and repair-pathway choice are the two numbers every CRISPR guide design pipeline is built to optimize.