CRISPR Guide RNA Designer
Scan for PAMs, score guides, and find the cut site
Paste a target sequence, scan both strands for NGG PAM sites, score every candidate guide against transparent design rules, and visualize the double-strand break.
The takeaway
Cas9 finds the PAM before it ever checks your guide. No PAM, no cut, which is the first and hardest constraint in genome editing.
GC-rich human locus; contains the widely-used EMX1 benchmark site (GTCACCTCCAATGACTAGGG + TGG).
Cas9 nicks both strands at the same position, producing a blunt double-strand break 3 bp upstream of the PAM, between protospacer positions 17 and 18. Coordinates are 1-based on the top strand you supplied. This guide targets the bottom strand, so its protospacer reads 5'→3' right-to-left relative to the sequence above, and its PAM sits to the left of the protospacer footprint.
- +50BaselineEvery candidate starts at 50 and is moved by the rules below.
- +25GC content 55%40–60% is the sweet spot: enough G/C for stable RNA:DNA duplex formation, not so much that the guide sticks to unintended sites.
- +8G at position 20The base immediately 5' of the PAM sits in the seed region. Empirically, a G here is associated with higher cleavage efficiency.
No other 20-mer in this sequence comes within 3 mismatches of the protospacer, the guide is unique within the sequence you supplied.
Be honest about this: the count above searches only the few hundred bases you pasted in. A real off-target assessment aligns the protospacer against the entire 3.1 billion-base genome (with a PAM-aware index such as Cas-OFFinder or CRISPOR), weights mismatches by position, mismatches in the PAM-proximal seed are far more disruptive than distal ones, and is ultimately confirmed empirically with GUIDE-seq or CIRCLE-seq. Treat this panel as a lower bound and a teaching aid, never as clearance to edit.
- The guide finds the target. A 20 nt spacer inside the sgRNA base-pairs with a matching 20 nt protospacer in the genome.
- The PAM licenses the cut. Cas9 only engages if the DNA immediately 3' of the protospacer reads NGG. No PAM, no cut, which is also why the bacterium does not shred its own CRISPR array.
- Cas9 makes a blunt double-strand break between positions 17 and 18 of the protospacer, exactly 3 bp upstream of the PAM.
- The cell repairs the break , and that repair, not the cut, is what edits the genome.
Non-homologous end joining glues the ends back together sloppily, inserting or deleting a few bases. If the indel is not a multiple of 3 the reading frame shifts, a premature stop appears, and the gene is destroyed. Cheap, efficient, imprecise.
If you supply a donor template with homology arms, homology- directed repair copies your sequence into the break, letting you correct a point mutation or knock a gene in. Precise, but only active in dividing cells and far less efficient than NHEJ.
This is a transparent, hand-written heuristic, not a machine-learned model. It is meant to teach you why guides fail. Real design tools (Doench 2016 / Rule Set 2, DeepSpCas9) are trained on thousands of measured cleavage events and will disagree with the numbers here.
- Baseline 50
Every candidate starts in the middle of the scale.
- GC content ±25
40–60% ideal. Below 30% the guide binds too weakly; above 80% it binds so tightly it tolerates mismatches and cuts off-target.
- Poly-T −30
TTTT is the Pol III terminator, the U6 promoter would abort transcription of the sgRNA itself.
- Homopolymers −6 / −15
Runs of 4 (or 5+) identical bases cause synthesis slippage and poor folding.
- Position 20 ±8
G adjacent to the PAM helps; C adjacent to the PAM hurts.
- Hairpins −5 each (max −15)
Self-complementary segments make the sgRNA fold on itself instead of loading into Cas9.
Cas9 only cuts where the ~20-nucleotide guide RNA matches the target, immediately next to a PAM sequence.
Image: Mariuswalter, Wikimedia Commons (CC BY-SA 4.0)